doi
stringlengths 9
10
| chunk-id
stringlengths 1
4
| chunk
stringlengths 1
1.57k
| id
stringlengths 9
10
| title
stringlengths 10
127
| summary
stringlengths 581
1.92k
| source
stringlengths 30
31
| authors
sequence | categories
sequence | comment
stringclasses 28
values | journal_ref
stringclasses 1
value | primary_category
stringclasses 6
values | published
stringlengths 8
8
| updated
stringlengths 8
8
| references
list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1806.01261 | 568 | sha1_base64="EsSika+7GeuHZg/021QVL5D37Tg=">AAAB7nicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoAcPAS8eI5gHJGuYnfQmQ2YfzPQKIeQjvHhQxKvf482/cZLsQRMLGoqqbrq7glRJQ6777RTW1jc2t4rbpZ3dvf2D8uFR0ySZFtgQiUp0O+AGlYyxQZIUtlONPAoUtoLR7cxvPaE2MokfaJyiH/FBLEMpOFmp1Q2Q+CP2yhW36s7BVomXkwrkqPfKX91+IrIIYxKKG9Px3JT8CdckhcJpqZsZTLkY8 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 569 | PfKX91+IrIIYxKKG9Px3JT8CdckhcJpqZsZTLkY8QF2LI15hMafzM+dsjOr9FmYaFsxsbn6e2LCI2PGUWA7I05Ds+zNxP+8TkbhtT+RcZoRxmKxKMwUo4TNfmd9qVGQGlvChZb2ViaGXHNBNqGSDcFbfnmVNC+qnlv17i8rtZs8jiKcwCmcgwdXUIM7qEMDBIzgGV7hzUmdF+fd+Vi0Fpx85hj+wPn8ATfaj3U=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 570 | sha1_base64="EsSika+7GeuHZg/021QVL5D37Tg=">AAAB7nicbVDLSgNBEOyNrxhfUY9eBoPgKeyKoAcPAS8eI5gHJGuYnfQmQ2YfzPQKIeQjvHhQxKvf482/cZLsQRMLGoqqbrq7glRJQ6777RTW1jc2t4rbpZ3dvf2D8uFR0ySZFtgQiUp0O+AGlYyxQZIUtlONPAoUtoLR7cxvPaE2MokfaJyiH/FBLEMpOFmp1Q2Q+CP2yhW36s7BVomXkwrkqPfKX91+IrIIYxKKG9Px3JT8CdckhcJpqZsZTLkY8 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 571 | PfKX91+IrIIYxKKG9Px3JT8CdckhcJpqZsZTLkY8QF2LI15hMafzM+dsjOr9FmYaFsxsbn6e2LCI2PGUWA7I05Ds+zNxP+8TkbhtT+RcZoRxmKxKMwUo4TNfmd9qVGQGlvChZb2ViaGXHNBNqGSDcFbfnmVNC+qnlv17i8rtZs8jiKcwCmcgwdXUIM7qEMDBIzgGV7hzUmdF+fd+Vi0Fpx85hj+wPn8ATfaj3U=</latexit>⇢e!v | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 572 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 573 | sha1_base64="s3/Cw/iD/Ic9TAit26LWmPV1hK0=">AAAB/nicbVBNS8NAEN3Ur1q/ouLJy2IRPJVEBD0WvXisYD+giWWznTRLN9mwu6mUUPCvePGgiFd/hzf/jds2B219MPB4b4aZeUHKmdKO822VVlbX1jfKm5Wt7Z3dPXv/oKVEJik0qeBCdgKigLMEmpppDp1UAokDDu1geDP12yOQionkXo9T8GMySFjIKNFG6tlHnozEQw6eZINIEynFIx5NenbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqP | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 574 | nbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqPydSM8phUvEyBSmhQzKArqEJiUH5+ez8CT41Sh+HQppKNJ6pvydyEis1jgPTGRMdqUVvKv7ndTMdXvk5S9JMQ0Lni8KMYy3wNAvcZxKo5mNDCJXM3IppRCSh2iRWMSG4iy8vk9Z5zXVq7t1FtX5dxFFGx+gEnSEXXaI6ukUN1EQU5egZvaI368l6sd6tj3lrySpmDtEfWJ8/7hmWGA==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 575 | sha1_base64="s3/Cw/iD/Ic9TAit26LWmPV1hK0=">AAAB/nicbVBNS8NAEN3Ur1q/ouLJy2IRPJVEBD0WvXisYD+giWWznTRLN9mwu6mUUPCvePGgiFd/hzf/jds2B219MPB4b4aZeUHKmdKO822VVlbX1jfKm5Wt7Z3dPXv/oKVEJik0qeBCdgKigLMEmpppDp1UAokDDu1geDP12yOQionkXo9T8GMySFjIKNFG6tlHnozEQw6eZINIEynFIx5NenbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqP | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 576 | nbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqPydSM8phUvEyBSmhQzKArqEJiUH5+ez8CT41Sh+HQppKNJ6pvydyEis1jgPTGRMdqUVvKv7ndTMdXvk5S9JMQ0Lni8KMYy3wNAvcZxKo5mNDCJXM3IppRCSh2iRWMSG4iy8vk9Z5zXVq7t1FtX5dxFFGx+gEnSEXXaI6ukUN1EQU5egZvaI368l6sd6tj3lrySpmDtEfWJ8/7hmWGA==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 577 | sha1_base64="s3/Cw/iD/Ic9TAit26LWmPV1hK0=">AAAB/nicbVBNS8NAEN3Ur1q/ouLJy2IRPJVEBD0WvXisYD+giWWznTRLN9mwu6mUUPCvePGgiFd/hzf/jds2B219MPB4b4aZeUHKmdKO822VVlbX1jfKm5Wt7Z3dPXv/oKVEJik0qeBCdgKigLMEmpppDp1UAokDDu1geDP12yOQionkXo9T8GMySFjIKNFG6tlHnozEQw6eZINIEynFIx5NenbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqP | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 578 | nbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqPydSM8phUvEyBSmhQzKArqEJiUH5+ez8CT41Sh+HQppKNJ6pvydyEis1jgPTGRMdqUVvKv7ndTMdXvk5S9JMQ0Lni8KMYy3wNAvcZxKo5mNDCJXM3IppRCSh2iRWMSG4iy8vk9Z5zXVq7t1FtX5dxFFGx+gEnSEXXaI6ukUN1EQU5egZvaI368l6sd6tj3lrySpmDtEfWJ8/7hmWGA==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 579 | sha1_base64="s3/Cw/iD/Ic9TAit26LWmPV1hK0=">AAAB/nicbVBNS8NAEN3Ur1q/ouLJy2IRPJVEBD0WvXisYD+giWWznTRLN9mwu6mUUPCvePGgiFd/hzf/jds2B219MPB4b4aZeUHKmdKO822VVlbX1jfKm5Wt7Z3dPXv/oKVEJik0qeBCdgKigLMEmpppDp1UAokDDu1geDP12yOQionkXo9T8GMySFjIKNFG6tlHnozEQw6eZINIEynFIx5NenbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqP | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 580 | nbVqTkz4GXiFqSKCjR69pfXFzSLIdGUE6W6rpNqPydSM8phUvEyBSmhQzKArqEJiUH5+ez8CT41Sh+HQppKNJ6pvydyEis1jgPTGRMdqUVvKv7ndTMdXvk5S9JMQ0Lni8KMYy3wNAvcZxKo5mNDCJXM3IppRCSh2iRWMSG4iy8vk9Z5zXVq7t1FtX5dxFFGx+gEnSEXXaI6ukUN1EQU5egZvaI368l6sd6tj3lrySpmDtEfWJ8/7hmWGA==</latexit> e | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 581 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 582 | sha1_base64="gRKFy+QFytmwqWy0cvo5FmmPz8I=">AAAB7XicbVA9SwNBEJ3zM8avqGBjsxgEq3Bno4VFwMYygrkEkjPubeaSNXu3x+6eEo78BxsLRWz9P3b+GzcfhSY+GHi8N8PMvDAVXBvX/XaWlldW19YLG8XNre2d3dLevq9lphjWmRRSNUOqUfAE64Ybgc1UIY1DgY1wcDX2G4+oNJfJrRmmGMS0l/CIM2qs5LfTPr/DTqnsVtwJyCLxZqRcPfSf7gGg1il9tbuSZTEmhgmqdctzUxPkVBnOBI6K7 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 583 | fSf7gGg1il9tbuSZTEmhgmqdctzUxPkVBnOBI6K7UxjStmA9rBlaUJj1EE+uXZETqzSJZFUthJDJurviZzGWg/j0HbG1PT1vDcW//NamYkugpwnaWYwYdNFUSaIkWT8OulyhcyIoSWUKW5vJaxPFWXGBlS0IXjzLy8S/6ziuRXvxqZxCVMU4AiO4RQ8OIcqXEMN6sDgAZ7hFd4c6bw4787HtHXJmc0cwB84nz+QDpCP</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 584 | sha1_base64="74MJShuZzxGyM2nLY3EY87InuhI=">AAAB7XicbVC7SgNBFJ2NrxhfUcHGZjAIVmHXRguLgI1lBLMJJGuYndxNxszODjOzyrLkH2wsFNHS/7HzA/wPJ49CEw9cOJxzL/feE0rOtHHdL6ewtLyyulZcL21sbm3vlHf3fJ2kikKDJjxRrZBo4ExAwzDDoSUVkDjk0AyHl2O/eQ9Ks0TcmExCEJO+YBGjxFjJ78gBu4VuueJW3QnwIvFmpFI78B9k9v1e75Y/O72EpjEIQznRuu250gQ5UYZRD | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 585 | FI78B9k9v1e75Y/O72EpjEIQznRuu250gQ5UYZRDqNSJ9UgCR2SPrQtFSQGHeSTa0f42Co9HCXKljB4ov6eyEmsdRaHtjMmZqDnvbH4n9dOTXQe5EzI1ICg00VRyrFJ8Ph13GMKqOGZJYQqZm/FdEAUocYGVLIhePMvLxL/tOq5Ve/apnGBpiiiQ3SETpCHzlANXaE6aiCK7tAjekYvTuI8Oa/O27S14Mxm9tEfOB8/TZOSnA==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 586 | sha1_base64="74MJShuZzxGyM2nLY3EY87InuhI=">AAAB7XicbVC7SgNBFJ2NrxhfUcHGZjAIVmHXRguLgI1lBLMJJGuYndxNxszODjOzyrLkH2wsFNHS/7HzA/wPJ49CEw9cOJxzL/feE0rOtHHdL6ewtLyyulZcL21sbm3vlHf3fJ2kikKDJjxRrZBo4ExAwzDDoSUVkDjk0AyHl2O/eQ9Ks0TcmExCEJO+YBGjxFjJ78gBu4VuueJW3QnwIvFmpFI78B9k9v1e75Y/O72EpjEIQznRuu250gQ5UYZRD | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 587 | FI78B9k9v1e75Y/O72EpjEIQznRuu250gQ5UYZRDqNSJ9UgCR2SPrQtFSQGHeSTa0f42Co9HCXKljB4ov6eyEmsdRaHtjMmZqDnvbH4n9dOTXQe5EzI1ICg00VRyrFJ8Ph13GMKqOGZJYQqZm/FdEAUocYGVLIhePMvLxL/tOq5Ve/apnGBpiiiQ3SETpCHzlANXaE6aiCK7tAjekYvTuI8Oa/O27S14Mxm9tEfOB8/TZOSnA==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 588 | sha1_base64="pLq6KB/1S9uyUeWp/G4byg43mK0=">AAAB7XicbVA9SwNBEJ2LXzF+RS1tFoNgFe5stLAI2FhGMB+QnGFvM5es2ds9dveEEPIfbCwUsfX/2Plv3CRXaOKDgcd7M8zMi1LBjfX9b6+wtr6xuVXcLu3s7u0flA+PmkZlmmGDKaF0O6IGBZfYsNwKbKcaaRIJbEWjm5nfekJtuJL3dpximNCB5DFn1Dqp2U2H/AF75Ypf9ecgqyTISQVy1Hvlr25fsSxBaZmgxnQCP7XhhGrLmcBpqZsZTCkb0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 589 | Hvlr25fsSxBaZmgxnQCP7XhhGrLmcBpqZsZTCkb0QF2HJU0QRNO5tdOyZlT+iRW2pW0ZK7+npjQxJhxErnOhNqhWfZm4n9eJ7PxVTjhMs0sSrZYFGeCWEVmr5M+18isGDtCmebuVsKGVFNmXUAlF0Kw/PIqaV5UA78a3PmV2nUeRxFO4BTOIYBLqMEt1KEBDB7hGV7hzVPei/fufSxaC14+cwx/4H3+AIWDjww=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 590 | sha1_base64="m8MJ1M94ujO0d0COo5n2Dsol6rc=">AAAB53icbVC7SgNBFL0bXzG+opY2g0GwCrs2phAM2FhGMA9IFpmdvZsMmZ1dZmaFsKS0sbFQxNZP8Rfs/AZ/wsmj0MQDFw7nnMt9BKng2rjul1NYWV1b3yhulra2d3b3yvsHLZ1kimGTJSJRnYBqFFxi03AjsJMqpHEgsB0MryZ++x6V5om8NaMU/Zj2JY84o8ZKjbtyxa26U5Bl4s1J5fLj+4EAgM1/9sKEZTFKwwTVuuu5qfFzqgxnAselXqYxp | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 591 | 4EAgM1/9sKEZTFKwwTVuuu5qfFzqgxnAselXqYxpWxI+9i1VNIYtZ9P9xyTE6uEJEqULWnIVP3dkdNY61Ec2GRMzUAvehPxP6+bmajm51ymmUHJZoOiTBCTkMnRJOQKmREjSyhT3O5K2IAqyox9Tck+wVs8eZm0zqqeW/Vu3Er9AmYowhEcwyl4cA51uIYGNIFBCI/wDC8Od56cV+dtFi04855D+APn/Qd/sI8A</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 592 | sha1_base64="q1zM5ZsCNMJAtNUZI97B98ATlaA=">AAAB53icbVC7SgNBFL3rM8ZX1FKQwSBYhV0bLQQDNpYJmAckQWZn7yZDZmeXmVkhLCltbCwUsfUvbP0FO79BP8LJo9DEAxcO55zLffiJ4Nq47qezsLi0vLKaW8uvb2xubRd2dus6ThXDGotFrJo+1Si4xJrhRmAzUUgjX2DD71+O/MYtKs1jeW0GCXYi2pU85IwaK1VuCkW35I5B5ok3JcWL96+7g7fqt81/tIOYpRFKwwTVuuW5ielkVBnOBA7z7 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 593 | 6+7g7fqt81/tIOYpRFKwwTVuuW5ielkVBnOBA7z7VRjQlmfdrFlqaQR6k423nNIjqwSkDBWtqQhY/V3R0YjrQeRb5MRNT09643E/7xWasKzTsZlkhqUbDIoTAUxMRkdTQKukBkxsIQyxe2uhPWooszY1+TtE7zZk+dJ/aTkuSWv6hbL5zBBDvbhEI7Bg1MowxVUoAYMAriHR3hyuPPgPDsvk+iCM+3Zgz9wXn8AGkeQ8w==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 594 | sha1_base64="q1zM5ZsCNMJAtNUZI97B98ATlaA=">AAAB53icbVC7SgNBFL3rM8ZX1FKQwSBYhV0bLQQDNpYJmAckQWZn7yZDZmeXmVkhLCltbCwUsfUvbP0FO79BP8LJo9DEAxcO55zLffiJ4Nq47qezsLi0vLKaW8uvb2xubRd2dus6ThXDGotFrJo+1Si4xJrhRmAzUUgjX2DD71+O/MYtKs1jeW0GCXYi2pU85IwaK1VuCkW35I5B5ok3JcWL96+7g7fqt81/tIOYpRFKwwTVuuW5ielkVBnOBA7z7 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 595 | 6+7g7fqt81/tIOYpRFKwwTVuuW5ielkVBnOBA7z7VRjQlmfdrFlqaQR6k423nNIjqwSkDBWtqQhY/V3R0YjrQeRb5MRNT09643E/7xWasKzTsZlkhqUbDIoTAUxMRkdTQKukBkxsIQyxe2uhPWooszY1+TtE7zZk+dJ/aTkuSWv6hbL5zBBDvbhEI7Bg1MowxVUoAYMAriHR3hyuPPgPDsvk+iCM+3Zgz9wXn8AGkeQ8w==</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 596 | sha1_base64="ioxb3woZF1oAlTScqds23PrgiMY=">AAAB53icbVC7SgNBFL3rM8ZX1NJmMAhWYdZGC4uAjWUE84BkkdnZ2WTI7Owyc1cIS37AxkIRW3/Jzr9xkmyhiQcGDuecy9x7wkxJi5R+e2vrG5tb25Wd6u7e/sFh7ei4Y9PccNHmqUpNL2RWKKlFGyUq0cuMYEmoRDcc38787pMwVqb6ASeZCBI21DKWnKGTWo+1Om3QOcgq8UtShxIu/zWIUp4nQiNXzNq+TzMMCmZQciWm1UFuRcb4mA1F31HNE | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 597 | p4nQiNXzNq+TzMMCmZQciWm1UFuRcb4mA1F31HNEmGDYr7nlJw7JSJxatzTSObq74mCJdZOktAlE4Yju+zNxP+8fo7xdVBIneUoNF98FOeKYEpmR5NIGsFRTRxh3Ei3K+EjZhhHV03VleAvn7xKOpcNnzb8e1pv3pR1VOAUzuACfLiCJtxBC9rAIYJneIU3T3ov3rv3sYiueeXMCfyB9/kDCGmMcA==</latexit> v | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 598 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 599 | sha1_base64="HCiXjOq04H3f4Ed7vqyiRfd+2dI=">AAAB7XicbVA9SwNBEJ2LXzF+nQo2NotBsAp3NlpYBGwsI5hLIDnj3mYvWbO3e+zuRcKR/2BjoYit/8fOf+Pmo9DEBwOP92aYmRelnGnjed9OYWV1bX2juFna2t7Z3XP3DwItM0VonUguVTPCmnImaN0ww2kzVRQnEaeNaHA98RtDqjST4s6MUhomuCdYzAg2VgraaZ/dDztu2at4U6Bl4s9JuXoUPD0AQK3jfrW7kmQJFYZwrHXL91IT5lgZRjgdl | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 600 | XoUPD0AQK3jfrW7kmQJFYZwrHXL91IT5lgZRjgdl9qZpikmA9yjLUsFTqgO8+m1Y3RqlS6KpbIlDJqqvydynGg9SiLbmWDT14veRPzPa2UmvgxzJtLMUEFmi+KMIyPR5HXUZYoSw0eWYKKYvRWRPlaYGBtQyYbgL768TILziu9V/FubxhXMUIRjOIEz8OECqnADNagDgUd4hld4c6Tz4rw7H7PWgjOfOYQ/cD5/AKnSkKA=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 601 | sha1_base64="/voSHBGyFE5xYPVKPYM/GTIarLQ=">AAAB7XicbVC7SgNBFL3rM8ZXVLCxWQyCVdi10cIiYGMZwWwCyRpmJ7PJmNmZYWY2siz5BxsLRbT0f+z8AP/DyaPQxAMXDufcy733RJJRbTzvy1laXlldWy9sFDe3tnd2S3v7gRapwqSOBROqGSFNGOWkbqhhpCkVQUnESCMaXI39xpAoTQW/NZkkYYJ6nMYUI2OloC379G7YKZW9ijeBu0j8GSlXD4MHmX2/1zqlz3ZX4DQh3GCGtG75njRhjpShm | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 602 | SlXD4MHmX2/1zqlz3ZX4DQh3GCGtG75njRhjpShmJFRsZ1qIhEeoB5pWcpRQnSYT64duSdW6bqxULa4cSfq74kcJVpnSWQ7E2T6et4bi/95rdTEF2FOuUwN4Xi6KE6Za4Q7ft3tUkWwYZklCCtqb3VxHymEjQ2oaEPw519eJMFZxfcq/o1N4xKmKMARHMMp+HAOVbiGGtQBwz08wjO8OMJ5cl6dt2nrkjObOYA/cD5+AGdXkq0=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 603 | sha1_base64="/voSHBGyFE5xYPVKPYM/GTIarLQ=">AAAB7XicbVC7SgNBFL3rM8ZXVLCxWQyCVdi10cIiYGMZwWwCyRpmJ7PJmNmZYWY2siz5BxsLRbT0f+z8AP/DyaPQxAMXDufcy733RJJRbTzvy1laXlldWy9sFDe3tnd2S3v7gRapwqSOBROqGSFNGOWkbqhhpCkVQUnESCMaXI39xpAoTQW/NZkkYYJ6nMYUI2OloC379G7YKZW9ijeBu0j8GSlXD4MHmX2/1zqlz3ZX4DQh3GCGtG75njRhjpShm | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 604 | SlXD4MHmX2/1zqlz3ZX4DQh3GCGtG75njRhjpShmJFRsZ1qIhEeoB5pWcpRQnSYT64duSdW6bqxULa4cSfq74kcJVpnSWQ7E2T6et4bi/95rdTEF2FOuUwN4Xi6KE6Za4Q7ft3tUkWwYZklCCtqb3VxHymEjQ2oaEPw519eJMFZxfcq/o1N4xKmKMARHMMp+HAOVbiGGtQBwz08wjO8OMJ5cl6dt2nrkjObOYA/cD5+AGdXkq0=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 605 | sha1_base64="Fc8T4ygtia14k1z/CDji4ezWDqY=">AAAB7XicbVA9SwNBEJ3zM8avqKXNYhCswp2NFhYBG8sI5gOSM+xtNsmavd1jdy4QjvwHGwtFbP0/dv4bN8kVmvhg4PHeDDPzokQKi77/7a2tb2xubRd2irt7+weHpaPjhtWpYbzOtNSmFVHLpVC8jgIlbyWG0ziSvBmNbmd+c8yNFVo94CThYUwHSvQFo+ikRicZisdxt1T2K/4cZJUEOSlDjlq39NXpaZbGXCGT1Np24CcYZtSgYJJPi53U8oSyE | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 606 | lq39NXpaZbGXCGT1Np24CcYZtSgYJJPi53U8oSyER3wtqOKxtyG2fzaKTl3So/0tXGlkMzV3xMZja2dxJHrjCkO7bI3E//z2in2r8NMqCRFrthiUT+VBDWZvU56wnCGcuIIZUa4WwkbUkMZuoCKLoRg+eVV0risBH4luPfL1Zs8jgKcwhlcQABXUIU7qEEdGDzBM7zCm6e9F+/d+1i0rnn5zAn8gff5A59Hjx0=</latexit>V0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 607 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 608 | sha1_base64="gAQ7qdt3IKvK5oBqK3uN1PHYi1k=">AAAB6XicbVA9SwNBEJ2LXzF+RS1tFoNoFe4koIVFwMYyivmA5Ah7m7lkyd7esbsnhCP/wMZCEVv/kZ3/xk1yhSY+GHi8N8PMvCARXBvX/XYKa+sbm1vF7dLO7t7+QfnwqKXjVDFssljEqhNQjYJLbBpuBHYShTQKBLaD8e3Mbz+h0jyWj2aSoB/RoeQhZ9RY6aF13i9X3Ko7B1klXk4qkKPRL3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOs | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 609 | 3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOsWuppBFqP5tfOiVnVhmQMFa2pCFz9fdERiOtJ1FgOyNqRnrZm4n/ed3UhNd+xmWSGpRssShMBTExmb1NBlwhM2JiCWWK21sJG1FFmbHhlGwI3vLLq6R1WfXcqndfq9Rv8jiKcAKncAEeXEEd7qABTWAQwjO8wpszdl6cd+dj0Vpw8plj+APn8wcRSI0F</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 610 | sha1_base64="gAQ7qdt3IKvK5oBqK3uN1PHYi1k=">AAAB6XicbVA9SwNBEJ2LXzF+RS1tFoNoFe4koIVFwMYyivmA5Ah7m7lkyd7esbsnhCP/wMZCEVv/kZ3/xk1yhSY+GHi8N8PMvCARXBvX/XYKa+sbm1vF7dLO7t7+QfnwqKXjVDFssljEqhNQjYJLbBpuBHYShTQKBLaD8e3Mbz+h0jyWj2aSoB/RoeQhZ9RY6aF13i9X3Ko7B1klXk4qkKPRL3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOs | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 611 | 3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOsWuppBFqP5tfOiVnVhmQMFa2pCFz9fdERiOtJ1FgOyNqRnrZm4n/ed3UhNd+xmWSGpRssShMBTExmb1NBlwhM2JiCWWK21sJG1FFmbHhlGwI3vLLq6R1WfXcqndfq9Rv8jiKcAKncAEeXEEd7qABTWAQwjO8wpszdl6cd+dj0Vpw8plj+APn8wcRSI0F</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 612 | sha1_base64="gAQ7qdt3IKvK5oBqK3uN1PHYi1k=">AAAB6XicbVA9SwNBEJ2LXzF+RS1tFoNoFe4koIVFwMYyivmA5Ah7m7lkyd7esbsnhCP/wMZCEVv/kZ3/xk1yhSY+GHi8N8PMvCARXBvX/XYKa+sbm1vF7dLO7t7+QfnwqKXjVDFssljEqhNQjYJLbBpuBHYShTQKBLaD8e3Mbz+h0jyWj2aSoB/RoeQhZ9RY6aF13i9X3Ko7B1klXk4qkKPRL3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOs | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 613 | 3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOsWuppBFqP5tfOiVnVhmQMFa2pCFz9fdERiOtJ1FgOyNqRnrZm4n/ed3UhNd+xmWSGpRssShMBTExmb1NBlwhM2JiCWWK21sJG1FFmbHhlGwI3vLLq6R1WfXcqndfq9Rv8jiKcAKncAEeXEEd7qABTWAQwjO8wpszdl6cd+dj0Vpw8plj+APn8wcRSI0F</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 614 | sha1_base64="gAQ7qdt3IKvK5oBqK3uN1PHYi1k=">AAAB6XicbVA9SwNBEJ2LXzF+RS1tFoNoFe4koIVFwMYyivmA5Ah7m7lkyd7esbsnhCP/wMZCEVv/kZ3/xk1yhSY+GHi8N8PMvCARXBvX/XYKa+sbm1vF7dLO7t7+QfnwqKXjVDFssljEqhNQjYJLbBpuBHYShTQKBLaD8e3Mbz+h0jyWj2aSoB/RoeQhZ9RY6aF13i9X3Ko7B1klXk4qkKPRL3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOs | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 615 | 3/1BjFLI5SGCap113MT42dUGc4ETku9VGNC2ZgOsWuppBFqP5tfOiVnVhmQMFa2pCFz9fdERiOtJ1FgOyNqRnrZm4n/ed3UhNd+xmWSGpRssShMBTExmb1NBlwhM2JiCWWK21sJG1FFmbHhlGwI3vLLq6R1WfXcqndfq9Rv8jiKcAKncAEeXEEd7qABTWAQwjO8wpszdl6cd+dj0Vpw8plj+APn8wcRSI0F</latexit>Figure | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 616 | 5: NLNNs as GNs. A schematic showing how NLNNs (Wang et al., 2018c) are implemented | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 617 | by the eande!vunder the GN framework. Typically, NLNNs assume that di erent regions of
an image (or words in a sentence) correspond to nodes in a fully connected graph, and the attention
mechanism de nes a weighted sum over nodes during the aggregation step.
the readout function, R, plays the role of the GN's u, but does not take uorE0as input,
and thus an analog to the GN's e!uis not required;
dmaster serves a roughly similar purpose to the GN's u, but is de ned as an extra node
connected to all others, and thus does not in
uence the edge and global updates directly. It
can then be represented in the GN's V.
Figure 4c shows how an MPNN is structured, according to the GN framework. For details and
various MPNN architectures, see the Appendix.
4.2.2 Non-local neural networks (NLNN)
Wang et al. (2018c)'s NLNN, which uni es various \intra-/self-/vertex-/graph-attention" approaches
(Lin et al., 2017; Vaswani et al., 2017; Hoshen, 2017; Veli ckovi c et al., 2018; Shaw et al., 2018), | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 618 | (Lin et al., 2017; Vaswani et al., 2017; Hoshen, 2017; Veli ckovi c et al., 2018; Shaw et al., 2018),
can also be translated into the GN formalism. The label \attention" refers to how the nodes are
updated: each node update is based on a weighted sum of (some function of) the node attributes of
its neighbors, where the weight between a node and one of its neighbors is computed by a scalar
pairwise function between their attributes (and then normalized across neighbors). The published
NLNN formalism does not explicitly include edges, and instead computes pairwise attention weights
between all nodes. But various NLNN-compliant models, such as the vertex attention interaction
network (Hoshen, 2017) and graph attention network (Veli ckovi c et al., 2018), are able to handle
explicit edges by e ectively setting to zero the weights between nodes which do not share an edge.
As Figures 4d and 5 illustrate, the eis factored into the scalar pairwise-interaction function
which returns the unnormalized attention term, denoted e(vrk;vsk)=a0
k, and a vector-valued
non-pairwise term, denoted e(vsk)=b0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 619 | k, and a vector-valued
non-pairwise term, denoted e(vsk)=b0
k. In thee!vaggregation, the a0
kterms are normalized
across each receiver's edges, b0
k, and elementwise summed:
e(ek;vrk;vsk;u):=fe(vrk;vsk) = ( e(vrk;vsk); e(vsk)) = (a0
k;b0
k) =e0
k
v e0
i;vi;u:=fv(e0
i)
e!v
E0
i:=1P
fk:rk=iga0
kX
fk:rk=iga0
kb0
k
In the NLNN paper's terminology (see Wang et al. (2018c), pages 2-4):
theirfplays the role of the above ,
17
theirgplays the role of the above .
This formulation may be helpful for focusing only on those interactions which are most relevant for
the downstream task, especially when the input entities were a set, from which a graph was formed
by adding all possible edges between them. | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 620 | the downstream task, especially when the input entities were a set, from which a graph was formed
by adding all possible edges between them.
Vaswani et al. (2017)'s multi-headed self-attention mechanism adds an interesting feature, where
the eande!vare implemented by a parallel set of functions, whose results are concatenated
together as the nal step of e!v. This can be interpreted as using typed edges, where the di erent
types index into di erent ecomponent functions, analogous to Li et al. (2016).
For details and various NLNN architectures, see the Appendix.
4.2.3 Other graph network variants
The full GN (Equation 2) can be used to predict a full graph, or any subset of (u0;V0;E0), as
outlined in Section 4.1.1. For example, to predict a global property of a graph, V0andE0can just
be ignored. Similarly, if global, node, or edge attributes are unspeci ed in the inputs, those vectors
can be zero-length, i.e., not taken as explicit input arguments. The same idea applies for other GN
variants which do not use the full set of mapping ( ) and reduction ( ) functions. For instance, | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 621 | variants which do not use the full set of mapping ( ) and reduction ( ) functions. For instance,
Interaction Networks (Battaglia et al., 2016; Watters et al., 2017) and the Neural Physics Engine
(Chang et al., 2017) use a full GN but for the absence of the global to update the edge properties
(see Appendix for details).
Various models, including CommNet (Sukhbaatar et al., 2016), structure2vec (Dai et al., 2016)
(in the version of (Dai et al., 2017)), and Gated Graph Sequence Neural Networks (Li et al., 2016)
have used a ewhich does not directly compute pairwise interactions, but instead ignore the receiver
node, operating only on the sender node and in some cases an edge attribute. This can be expressed
by implementations of ewith the following signatures, such as:
e(ek;vrk;vsk;u):=fe(vsk)
or e(ek;vrk;vsk;u):=vsk+fe(ek)
or e(ek;vrk;vsk;u):=fe(ek;vsk):
See the Appendix for further details.
Relation Networks (Raposo et al., 2017; Santoro et al., 2017) bypass the node update entirely | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 622 | See the Appendix for further details.
Relation Networks (Raposo et al., 2017; Santoro et al., 2017) bypass the node update entirely
and predict the global output from pooled edge information directly (see also Figure 4e),
e(ek;vrk;vsk;u):=fe(vrk;vsk) = NNe([vrk;vsk])
u
e0; v0;u:=fu
e0
= NNu
e0
e!u
E0:= =X
ke0
k
Deep Sets (Zaheer et al., 2017) bypass the edges update completely and predict the global output
from pooled nodes information directly (Figure 4f),
v(ei;vi;u):=fv(vi;u) = NNv([vi;u])
u
e0; v0;u:=fu
v0
= NNu
v0
v!u
V0:= =X
iv0
i
PointNet (Qi et al., 2017) use similar update rule, with a max-aggregation for v!uand a two-step
node update.
18
GM | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 623 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 624 | sha1_base64="vjTCpRgsPEJfhljVzwQb7AFhV5c=">AAAB6nicbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYKGNENF8QHKEvc1csmRv79jdE8KRn2BjoYitv8jOf+MmuUITHww83pthZl6QCK6N6347hZXVtfWN4mZpa3tnd6+8f9DUcaoYNlgsYtUOqEbBJTYMNwLbiUIaBQJbweh66reeUGkey0czTtCP6EDykDNqrPRw07vrlStu1Z2BLBMvJxXIUe+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsR | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 625 | e+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsRAfYsVTSCLWfzU6dkBOr9EkYK1vSkJn6eyKjkdbjKLCdETVDvehNxf+8TmrCSz/jMkkNSjZfFKaCmJhM/yZ9rpAZMbaEMsXtrYQNqaLM2HRKNgRv8eVl0jyrem7Vuz+v1K7yOIpwBMdwCh5cQA1uoQ4NYDCAZ3iFN0c4L8678zFvLTj5zCH8gfP5A+mdjYU=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 626 | sha1_base64="vjTCpRgsPEJfhljVzwQb7AFhV5c=">AAAB6nicbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYKGNENF8QHKEvc1csmRv79jdE8KRn2BjoYitv8jOf+MmuUITHww83pthZl6QCK6N6347hZXVtfWN4mZpa3tnd6+8f9DUcaoYNlgsYtUOqEbBJTYMNwLbiUIaBQJbweh66reeUGkey0czTtCP6EDykDNqrPRw07vrlStu1Z2BLBMvJxXIUe+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsR | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 627 | e+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsRAfYsVTSCLWfzU6dkBOr9EkYK1vSkJn6eyKjkdbjKLCdETVDvehNxf+8TmrCSz/jMkkNSjZfFKaCmJhM/yZ9rpAZMbaEMsXtrYQNqaLM2HRKNgRv8eVl0jyrem7Vuz+v1K7yOIpwBMdwCh5cQA1uoQ4NYDCAZ3iFN0c4L8678zFvLTj5zCH8gfP5A+mdjYU=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 628 | sha1_base64="vjTCpRgsPEJfhljVzwQb7AFhV5c=">AAAB6nicbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYKGNENF8QHKEvc1csmRv79jdE8KRn2BjoYitv8jOf+MmuUITHww83pthZl6QCK6N6347hZXVtfWN4mZpa3tnd6+8f9DUcaoYNlgsYtUOqEbBJTYMNwLbiUIaBQJbweh66reeUGkey0czTtCP6EDykDNqrPRw07vrlStu1Z2BLBMvJxXIUe+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsR | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 629 | e+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsRAfYsVTSCLWfzU6dkBOr9EkYK1vSkJn6eyKjkdbjKLCdETVDvehNxf+8TmrCSz/jMkkNSjZfFKaCmJhM/yZ9rpAZMbaEMsXtrYQNqaLM2HRKNgRv8eVl0jyrem7Vuz+v1K7yOIpwBMdwCh5cQA1uoQ4NYDCAZ3iFN0c4L8678zFvLTj5zCH8gfP5A+mdjYU=</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 630 | sha1_base64="vjTCpRgsPEJfhljVzwQb7AFhV5c=">AAAB6nicbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYKGNENF8QHKEvc1csmRv79jdE8KRn2BjoYitv8jOf+MmuUITHww83pthZl6QCK6N6347hZXVtfWN4mZpa3tnd6+8f9DUcaoYNlgsYtUOqEbBJTYMNwLbiUIaBQJbweh66reeUGkey0czTtCP6EDykDNqrPRw07vrlStu1Z2BLBMvJxXIUe+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsR | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 631 | e+Vv7r9mKURSsME1brjuYnxM6oMZwInpW6qMaFsRAfYsVTSCLWfzU6dkBOr9EkYK1vSkJn6eyKjkdbjKLCdETVDvehNxf+8TmrCSz/jMkkNSjZfFKaCmJhM/yZ9rpAZMbaEMsXtrYQNqaLM2HRKNgRv8eVl0jyrem7Vuz+v1K7yOIpwBMdwCh5cQA1uoQ4NYDCAZ3iFN0c4L8678zFvLTj5zCH8gfP5A+mdjYU=</latexit>GN1 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 632 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 633 | sha1_base64="oAmr7/S238q10w2wEvXkfEGmAr8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIUBcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72d9r1+uuFV3DrRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwO | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 634 | rRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwOiv1Ek1jTMZ4SLuWSiyo9tN56Bk6s8oAhZGyTxo0V39vpFhoPRWBncxC6mUvE//zuokJr/yUyTgxVJLFoTDhyEQoawANmKLE8KklmChmsyIywgoTY3sq2RK85S+vktZF1XOr3sNlpX6d11GEEziFc/CgBnW4gwY0gcATPMMrvDkT58V5dz4WowUn3zmGP3A+fwCg3ZH4</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 635 | sha1_base64="oAmr7/S238q10w2wEvXkfEGmAr8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIUBcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72d9r1+uuFV3DrRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwO | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 636 | rRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwOiv1Ek1jTMZ4SLuWSiyo9tN56Bk6s8oAhZGyTxo0V39vpFhoPRWBncxC6mUvE//zuokJr/yUyTgxVJLFoTDhyEQoawANmKLE8KklmChmsyIywgoTY3sq2RK85S+vktZF1XOr3sNlpX6d11GEEziFc/CgBnW4gwY0gcATPMMrvDkT58V5dz4WowUn3zmGP3A+fwCg3ZH4</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 637 | sha1_base64="oAmr7/S238q10w2wEvXkfEGmAr8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIUBcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72d9r1+uuFV3DrRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwO | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 638 | rRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwOiv1Ek1jTMZ4SLuWSiyo9tN56Bk6s8oAhZGyTxo0V39vpFhoPRWBncxC6mUvE//zuokJr/yUyTgxVJLFoTDhyEQoawANmKLE8KklmChmsyIywgoTY3sq2RK85S+vktZF1XOr3sNlpX6d11GEEziFc/CgBnW4gwY0gcATPMMrvDkT58V5dz4WowUn3zmGP3A+fwCg3ZH4</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 639 | sha1_base64="oAmr7/S238q10w2wEvXkfEGmAr8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIUBcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72d9r1+uuFV3DrRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwO | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 640 | rRKvJxUIEejX/7qDSKSCCoN4VjrrufGxk+xMoxwOiv1Ek1jTMZ4SLuWSiyo9tN56Bk6s8oAhZGyTxo0V39vpFhoPRWBncxC6mUvE//zuokJr/yUyTgxVJLFoTDhyEQoawANmKLE8KklmChmsyIywgoTY3sq2RK85S+vktZF1XOr3sNlpX6d11GEEziFc/CgBnW4gwY0gcATPMMrvDkT58V5dz4WowUn3zmGP3A+fwCg3ZH4</latexit>GN2 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 641 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 642 | sha1_base64="pet508CCa1uIZM8cv8xqNGylB9w=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqswUQRcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72f9Wr9ccavuHGiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhd | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 643 | GiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhdFbqJZrGmIzxkHYtlVhQ7afz0DN0ZpUBCiNlnzRorv7eSLHQeioCO5mF1MteJv7ndRMTXvkpk3FiqCSLQ2HCkYlQ1gAaMEWJ4VNLMFHMZkVkhBUmxvZUsiV4y19eJa1a1XOr3sNFpX6d11GEEziFc/DgEupwBw1oAoEneIZXeHMmzovz7nwsRgtOvnMMf+B8/gCiYZH5</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 644 | sha1_base64="pet508CCa1uIZM8cv8xqNGylB9w=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqswUQRcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72f9Wr9ccavuHGiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhd | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 645 | GiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhdFbqJZrGmIzxkHYtlVhQ7afz0DN0ZpUBCiNlnzRorv7eSLHQeioCO5mF1MteJv7ndRMTXvkpk3FiqCSLQ2HCkYlQ1gAaMEWJ4VNLMFHMZkVkhBUmxvZUsiV4y19eJa1a1XOr3sNFpX6d11GEEziFc/DgEupwBw1oAoEneIZXeHMmzovz7nwsRgtOvnMMf+B8/gCiYZH5</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 646 | sha1_base64="pet508CCa1uIZM8cv8xqNGylB9w=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqswUQRcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72f9Wr9ccavuHGiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhd | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 647 | GiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhdFbqJZrGmIzxkHYtlVhQ7afz0DN0ZpUBCiNlnzRorv7eSLHQeioCO5mF1MteJv7ndRMTXvkpk3FiqCSLQ2HCkYlQ1gAaMEWJ4VNLMFHMZkVkhBUmxvZUsiV4y19eJa1a1XOr3sNFpX6d11GEEziFc/DgEupwBw1oAoEneIZXeHMmzovz7nwsRgtOvnMMf+B8/gCiYZH5</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 648 | sha1_base64="pet508CCa1uIZM8cv8xqNGylB9w=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqswUQRcuCi50JRXsA9qhZNJMG5pkxiRTKEO/w40LRdz6Me78GzPtLLT1QOBwzr3ckxPEnGnjut9OYW19Y3OruF3a2d3bPygfHrV0lChCmyTikeoEWFPOJG0aZjjtxIpiEXDaDsY3md+eUKVZJB/NNKa+wEPJQkawsZLfE9iMlEhv72f9Wr9ccavuHGiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhd | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 649 | GiVeDmpQI5Gv/zVG0QkEVQawrHWXc+NjZ9iZRjhdFbqJZrGmIzxkHYtlVhQ7afz0DN0ZpUBCiNlnzRorv7eSLHQeioCO5mF1MteJv7ndRMTXvkpk3FiqCSLQ2HCkYlQ1gAaMEWJ4VNLMFHMZkVkhBUmxvZUsiV4y19eJa1a1XOr3sNFpX6d11GEEziFc/DgEupwBw1oAoEneIZXeHMmzovz7nwsRgtOvnMMf+B8/gCiYZH5</latexit>GNM | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 650 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 651 | sha1_base64="1uUQuLnXmq2FrQq5fvsHBbzm7v8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIoAsXBRe6USrYB7RDyaSZNjTJjEmmUIZ+hxsXirj1Y9z5N2baWWjrgcDhnHu5JyeIOdPGdb+dwsrq2vpGcbO0tb2zu1feP2jqKFGENkjEI9UOsKacSdowzHDajhXFIuC0FYyuM781pkqzSD6aSUx9gQeShYxgYyW/K7AZKpHe3E97d71yxa26M6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 652 | 6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE02mpm2gaYzLCA9qxVGJBtZ/OQk/RiVX6KIyUfdKgmfp7I8VC64kI7GQWUi96mfif10lMeOmnTMaJoZLMD4UJRyZCWQOozxQlhk8swUQxmxWRIVaYGNtTyZbgLX55mTTPqp5b9R7OK7WrvI4iHMExnIIHF1CDW6hDAwg8wTO8wpszdl6cd+djPlpw8p1D+APn8wfLTZIU</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 653 | sha1_base64="1uUQuLnXmq2FrQq5fvsHBbzm7v8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIoAsXBRe6USrYB7RDyaSZNjTJjEmmUIZ+hxsXirj1Y9z5N2baWWjrgcDhnHu5JyeIOdPGdb+dwsrq2vpGcbO0tb2zu1feP2jqKFGENkjEI9UOsKacSdowzHDajhXFIuC0FYyuM781pkqzSD6aSUx9gQeShYxgYyW/K7AZKpHe3E97d71yxa26M6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 654 | 6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE02mpm2gaYzLCA9qxVGJBtZ/OQk/RiVX6KIyUfdKgmfp7I8VC64kI7GQWUi96mfif10lMeOmnTMaJoZLMD4UJRyZCWQOozxQlhk8swUQxmxWRIVaYGNtTyZbgLX55mTTPqp5b9R7OK7WrvI4iHMExnIIHF1CDW6hDAwg8wTO8wpszdl6cd+djPlpw8p1D+APn8wfLTZIU</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 655 | sha1_base64="1uUQuLnXmq2FrQq5fvsHBbzm7v8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIoAsXBRe6USrYB7RDyaSZNjTJjEmmUIZ+hxsXirj1Y9z5N2baWWjrgcDhnHu5JyeIOdPGdb+dwsrq2vpGcbO0tb2zu1feP2jqKFGENkjEI9UOsKacSdowzHDajhXFIuC0FYyuM781pkqzSD6aSUx9gQeShYxgYyW/K7AZKpHe3E97d71yxa26M6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 656 | 6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE02mpm2gaYzLCA9qxVGJBtZ/OQk/RiVX6KIyUfdKgmfp7I8VC64kI7GQWUi96mfif10lMeOmnTMaJoZLMD4UJRyZCWQOozxQlhk8swUQxmxWRIVaYGNtTyZbgLX55mTTPqp5b9R7OK7WrvI4iHMExnIIHF1CDW6hDAwg8wTO8wpszdl6cd+djPlpw8p1D+APn8wfLTZIU</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 657 | sha1_base64="1uUQuLnXmq2FrQq5fvsHBbzm7v8=">AAAB9HicbVDLSgMxFL1TX7W+qi7dBIvgqsyIoAsXBRe6USrYB7RDyaSZNjTJjEmmUIZ+hxsXirj1Y9z5N2baWWjrgcDhnHu5JyeIOdPGdb+dwsrq2vpGcbO0tb2zu1feP2jqKFGENkjEI9UOsKacSdowzHDajhXFIuC0FYyuM781pkqzSD6aSUx9gQeShYxgYyW/K7AZKpHe3E97d71yxa26M6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE0 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 658 | 6Bl4uWkAjnqvfJXtx+RRFBpCMdadzw3Nn6KlWGE02mpm2gaYzLCA9qxVGJBtZ/OQk/RiVX6KIyUfdKgmfp7I8VC64kI7GQWUi96mfif10lMeOmnTMaJoZLMD4UJRyZCWQOozxQlhk8swUQxmxWRIVaYGNtTyZbgLX55mTTPqp5b9R7OK7WrvI4iHMExnIIHF1CDW6hDAwg8wTO8wpszdl6cd+djPlpw8p1D+APn8wfLTZIU</latexit>...<latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 659 | sha1_base64="Gj7yv98SlyD93Ghofp+NnyXvd2c=">AAAB7HicbVBNS8NAFHypX7V+VT16WSyCp5KIUA8eCl48VjBtoQ1ls920SzebsPsilNDf4MWDIl79Qd78N27bHLR1YGGYecO+N2EqhUHX/XZKG5tb2zvl3cre/sHhUfX4pG2STDPus0QmuhtSw6VQ3EeBkndTzWkcSt4JJ3dzv/PEtRGJesRpyoOYjpSIBKNoJb8/TNAMqjW37i5A1olXkBoUaA2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 660 | A2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI96zVNGYmyBfLDsjF1YZkijR9ikkC/V3IqexMdM4tJMxxbFZ9ebif14vw+gmyIVKM+SKLT+KMkkwIfPLyVBozlBOLaFMC7srYWOqKUPbT8WW4K2evE7aV3XPrXsP17XmbVFHGc7gHC7BgwY04R5a4AMDAc/wCm+Ocl6cd+djOVpyiswp/IHz+QPvmo68</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 661 | sha1_base64="Gj7yv98SlyD93Ghofp+NnyXvd2c=">AAAB7HicbVBNS8NAFHypX7V+VT16WSyCp5KIUA8eCl48VjBtoQ1ls920SzebsPsilNDf4MWDIl79Qd78N27bHLR1YGGYecO+N2EqhUHX/XZKG5tb2zvl3cre/sHhUfX4pG2STDPus0QmuhtSw6VQ3EeBkndTzWkcSt4JJ3dzv/PEtRGJesRpyoOYjpSIBKNoJb8/TNAMqjW37i5A1olXkBoUaA2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 662 | A2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI96zVNGYmyBfLDsjF1YZkijR9ikkC/V3IqexMdM4tJMxxbFZ9ebif14vw+gmyIVKM+SKLT+KMkkwIfPLyVBozlBOLaFMC7srYWOqKUPbT8WW4K2evE7aV3XPrXsP17XmbVFHGc7gHC7BgwY04R5a4AMDAc/wCm+Ocl6cd+djOVpyiswp/IHz+QPvmo68</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 663 | sha1_base64="Gj7yv98SlyD93Ghofp+NnyXvd2c=">AAAB7HicbVBNS8NAFHypX7V+VT16WSyCp5KIUA8eCl48VjBtoQ1ls920SzebsPsilNDf4MWDIl79Qd78N27bHLR1YGGYecO+N2EqhUHX/XZKG5tb2zvl3cre/sHhUfX4pG2STDPus0QmuhtSw6VQ3EeBkndTzWkcSt4JJ3dzv/PEtRGJesRpyoOYjpSIBKNoJb8/TNAMqjW37i5A1olXkBoUaA2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 664 | A2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI96zVNGYmyBfLDsjF1YZkijR9ikkC/V3IqexMdM4tJMxxbFZ9ebif14vw+gmyIVKM+SKLT+KMkkwIfPLyVBozlBOLaFMC7srYWOqKUPbT8WW4K2evE7aV3XPrXsP17XmbVFHGc7gHC7BgwY04R5a4AMDAc/wCm+Ocl6cd+djOVpyiswp/IHz+QPvmo68</latexit><latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 665 | sha1_base64="Gj7yv98SlyD93Ghofp+NnyXvd2c=">AAAB7HicbVBNS8NAFHypX7V+VT16WSyCp5KIUA8eCl48VjBtoQ1ls920SzebsPsilNDf4MWDIl79Qd78N27bHLR1YGGYecO+N2EqhUHX/XZKG5tb2zvl3cre/sHhUfX4pG2STDPus0QmuhtSw6VQ3EeBkndTzWkcSt4JJ3dzv/PEtRGJesRpyoOYjpSIBKNoJb8/TNAMqjW37i5A1olXkBoUaA2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 666 | A2qXzbHspgrZJIa0/PcFIOcahRM8lmlnxmeUjahI96zVNGYmyBfLDsjF1YZkijR9ikkC/V3IqexMdM4tJMxxbFZ9ebif14vw+gmyIVKM+SKLT+KMkkwIfPLyVBozlBOLaFMC7srYWOqKUPbT8WW4K2evE7aV3XPrXsP17XmbVFHGc7gHC7BgwY04R5a4AMDAc/wCm+Ocl6cd+djOVpyiswp/IHz+QPvmo68</latexit>G1 | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
1806.01261 | 667 | <latexit | 1806.01261 | Relational inductive biases, deep learning, and graph networks | Artificial intelligence (AI) has undergone a renaissance recently, making
major progress in key domains such as vision, language, control, and
decision-making. This has been due, in part, to cheap data and cheap compute
resources, which have fit the natural strengths of deep learning. However, many
defining characteristics of human intelligence, which developed under much
different pressures, remain out of reach for current approaches. In particular,
generalizing beyond one's experiences--a hallmark of human intelligence from
infancy--remains a formidable challenge for modern AI.
The following is part position paper, part review, and part unification. We
argue that combinatorial generalization must be a top priority for AI to
achieve human-like abilities, and that structured representations and
computations are key to realizing this objective. Just as biology uses nature
and nurture cooperatively, we reject the false choice between
"hand-engineering" and "end-to-end" learning, and instead advocate for an
approach which benefits from their complementary strengths. We explore how
using relational inductive biases within deep learning architectures can
facilitate learning about entities, relations, and rules for composing them. We
present a new building block for the AI toolkit with a strong relational
inductive bias--the graph network--which generalizes and extends various
approaches for neural networks that operate on graphs, and provides a
straightforward interface for manipulating structured knowledge and producing
structured behaviors. We discuss how graph networks can support relational
reasoning and combinatorial generalization, laying the foundation for more
sophisticated, interpretable, and flexible patterns of reasoning. As a
companion to this paper, we have released an open-source software library for
building graph networks, with demonstrations of how to use them in practice. | http://arxiv.org/pdf/1806.01261 | [
"Peter W. Battaglia",
"Jessica B. Hamrick",
"Victor Bapst",
"Alvaro Sanchez-Gonzalez",
"Vinicius Zambaldi",
"Mateusz Malinowski",
"Andrea Tacchetti",
"David Raposo",
"Adam Santoro",
"Ryan Faulkner",
"Caglar Gulcehre",
"Francis Song",
"Andrew Ballard",
"Justin Gilmer",
"George Dahl",
"Ashish Vaswani",
"Kelsey Allen",
"Charles Nash",
"Victoria Langston",
"Chris Dyer",
"Nicolas Heess",
"Daan Wierstra",
"Pushmeet Kohli",
"Matt Botvinick",
"Oriol Vinyals",
"Yujia Li",
"Razvan Pascanu"
] | [
"cs.LG",
"cs.AI",
"stat.ML"
] | null | null | cs.LG | 20180604 | 20181017 | [] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.