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 | 768 | <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 | 769 | sha1_base64="mvYjY6mgtt6w2Efm7YMP3XaOQ7I=">AAAB+XicbVDLSgMxFL1TX7W+Rl26CRbBVZkRQRcuCi50WcE+oB2GTJppQ5PMkGQKZeifuHGhiFv/xJ1/Y6adhVYPBA7n3EvOPVHKmTae9+VU1tY3Nreq27Wd3b39A/fwqKOTTBHaJglPVC/CmnImadsww2kvVRSLiNNuNLkt/O6UKs0S+WhmKQ0EHkkWM4KNlULXvQvzgcBmrETOZDqfh27da3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKM | 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 | 770 | 3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKMMLpvDbINE0xmeAR7VsqsaA6yBfJ5+jMKkMUJ8o+adBC/bmRY6H1TER2sgipV71C/M/rZya+DoqLMkMlWX4UZxyZBBU1oCFTlBg+swQTxWxWRMZYYWJsWTVbgr968l/SuWj4XsN/uKw3b8o6qnACp3AOPlxBE+6hBW0gMIUneIFXJ3eenTfnfTlaccqdY/gF5+MbP22UBw==</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 | 771 | sha1_base64="mvYjY6mgtt6w2Efm7YMP3XaOQ7I=">AAAB+XicbVDLSgMxFL1TX7W+Rl26CRbBVZkRQRcuCi50WcE+oB2GTJppQ5PMkGQKZeifuHGhiFv/xJ1/Y6adhVYPBA7n3EvOPVHKmTae9+VU1tY3Nreq27Wd3b39A/fwqKOTTBHaJglPVC/CmnImadsww2kvVRSLiNNuNLkt/O6UKs0S+WhmKQ0EHkkWM4KNlULXvQvzgcBmrETOZDqfh27da3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKM | 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 | 772 | 3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKMMLpvDbINE0xmeAR7VsqsaA6yBfJ5+jMKkMUJ8o+adBC/bmRY6H1TER2sgipV71C/M/rZya+DoqLMkMlWX4UZxyZBBU1oCFTlBg+swQTxWxWRMZYYWJsWTVbgr968l/SuWj4XsN/uKw3b8o6qnACp3AOPlxBE+6hBW0gMIUneIFXJ3eenTfnfTlaccqdY/gF5+MbP22UBw==</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 | 773 | sha1_base64="mvYjY6mgtt6w2Efm7YMP3XaOQ7I=">AAAB+XicbVDLSgMxFL1TX7W+Rl26CRbBVZkRQRcuCi50WcE+oB2GTJppQ5PMkGQKZeifuHGhiFv/xJ1/Y6adhVYPBA7n3EvOPVHKmTae9+VU1tY3Nreq27Wd3b39A/fwqKOTTBHaJglPVC/CmnImadsww2kvVRSLiNNuNLkt/O6UKs0S+WhmKQ0EHkkWM4KNlULXvQvzgcBmrETOZDqfh27da3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKM | 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 | 774 | 3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKMMLpvDbINE0xmeAR7VsqsaA6yBfJ5+jMKkMUJ8o+adBC/bmRY6H1TER2sgipV71C/M/rZya+DoqLMkMlWX4UZxyZBBU1oCFTlBg+swQTxWxWRMZYYWJsWTVbgr968l/SuWj4XsN/uKw3b8o6qnACp3AOPlxBE+6hBW0gMIUneIFXJ3eenTfnfTlaccqdY/gF5+MbP22UBw==</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 | 775 | sha1_base64="mvYjY6mgtt6w2Efm7YMP3XaOQ7I=">AAAB+XicbVDLSgMxFL1TX7W+Rl26CRbBVZkRQRcuCi50WcE+oB2GTJppQ5PMkGQKZeifuHGhiFv/xJ1/Y6adhVYPBA7n3EvOPVHKmTae9+VU1tY3Nreq27Wd3b39A/fwqKOTTBHaJglPVC/CmnImadsww2kvVRSLiNNuNLkt/O6UKs0S+WhmKQ0EHkkWM4KNlULXvQvzgcBmrETOZDqfh27da3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKM | 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 | 776 | 3gLoL/EL0kdSrRC93MwTEgmqDSEY637vpeaIMfKMMLpvDbINE0xmeAR7VsqsaA6yBfJ5+jMKkMUJ8o+adBC/bmRY6H1TER2sgipV71C/M/rZya+DoqLMkMlWX4UZxyZBBU1oCFTlBg+swQTxWxWRMZYYWJsWTVbgr968l/SuWj4XsN/uKw3b8o6qnACp3AOPlxBE+6hBW0gMIUneIFXJ3eenTfnfTlaccqdY/gF5+MbP22UBw==</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 | 777 | (b) Encode-process-decode | 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 | 778 | GNenc | 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 | 779 | <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 | 780 | sha1_base64="KZY5NxgXVEC/q8QikcRvAbEAxBs=">AAAB+nicbVDLSsNAFL3xWesr1aWbYBFclUQEXbgouNCVVLAPaEOYTKft0JlJmJkoJeZT3LhQxK1f4s6/cRKz0NYDA4dz7mXOPWHMqNKu+2UtLa+srq1XNqqbW9s7u3Ztr6OiRGLSxhGLZC9EijAqSFtTzUgvlgTxkJFuOL3M/e49kYpG4k7PYuJzNBZ0RDHSRgrs2oAjPZE8vbrJgpQInAV23W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iq | 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 | 781 | W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iqSlmJKsOEkVihKdoTPqGCsSJ8tMieuYcGWXojCJpntBOof7eSBFXasZDM5kHVfNeLv7n9RM9OvdTKuJE51cVH40S5ujIyXtwhlQSrNnMEIQlNVkdPEESYW3aqpoSvPmTF0nnpOG5De/2tN68KOuowAEcwjF4cAZNuIYWtAHDAzzBC7xaj9az9Wa9/4wuWeXOPvyB9fENyBqUTg==</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 | 782 | sha1_base64="KZY5NxgXVEC/q8QikcRvAbEAxBs=">AAAB+nicbVDLSsNAFL3xWesr1aWbYBFclUQEXbgouNCVVLAPaEOYTKft0JlJmJkoJeZT3LhQxK1f4s6/cRKz0NYDA4dz7mXOPWHMqNKu+2UtLa+srq1XNqqbW9s7u3Ztr6OiRGLSxhGLZC9EijAqSFtTzUgvlgTxkJFuOL3M/e49kYpG4k7PYuJzNBZ0RDHSRgrs2oAjPZE8vbrJgpQInAV23W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iq | 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 | 783 | W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iqSlmJKsOEkVihKdoTPqGCsSJ8tMieuYcGWXojCJpntBOof7eSBFXasZDM5kHVfNeLv7n9RM9OvdTKuJE51cVH40S5ujIyXtwhlQSrNnMEIQlNVkdPEESYW3aqpoSvPmTF0nnpOG5De/2tN68KOuowAEcwjF4cAZNuIYWtAHDAzzBC7xaj9az9Wa9/4wuWeXOPvyB9fENyBqUTg==</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 | 784 | sha1_base64="KZY5NxgXVEC/q8QikcRvAbEAxBs=">AAAB+nicbVDLSsNAFL3xWesr1aWbYBFclUQEXbgouNCVVLAPaEOYTKft0JlJmJkoJeZT3LhQxK1f4s6/cRKz0NYDA4dz7mXOPWHMqNKu+2UtLa+srq1XNqqbW9s7u3Ztr6OiRGLSxhGLZC9EijAqSFtTzUgvlgTxkJFuOL3M/e49kYpG4k7PYuJzNBZ0RDHSRgrs2oAjPZE8vbrJgpQInAV23W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iq | 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 | 785 | W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iqSlmJKsOEkVihKdoTPqGCsSJ8tMieuYcGWXojCJpntBOof7eSBFXasZDM5kHVfNeLv7n9RM9OvdTKuJE51cVH40S5ujIyXtwhlQSrNnMEIQlNVkdPEESYW3aqpoSvPmTF0nnpOG5De/2tN68KOuowAEcwjF4cAZNuIYWtAHDAzzBC7xaj9az9Wa9/4wuWeXOPvyB9fENyBqUTg==</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 | 786 | sha1_base64="KZY5NxgXVEC/q8QikcRvAbEAxBs=">AAAB+nicbVDLSsNAFL3xWesr1aWbYBFclUQEXbgouNCVVLAPaEOYTKft0JlJmJkoJeZT3LhQxK1f4s6/cRKz0NYDA4dz7mXOPWHMqNKu+2UtLa+srq1XNqqbW9s7u3Ztr6OiRGLSxhGLZC9EijAqSFtTzUgvlgTxkJFuOL3M/e49kYpG4k7PYuJzNBZ0RDHSRgrs2oAjPZE8vbrJgpQInAV23W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iq | 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 | 787 | W24BZxF4pWkDiVagf05GEY44URozJBSfc+NtZ8iqSlmJKsOEkVihKdoTPqGCsSJ8tMieuYcGWXojCJpntBOof7eSBFXasZDM5kHVfNeLv7n9RM9OvdTKuJE51cVH40S5ujIyXtwhlQSrNnMEIQlNVkdPEESYW3aqpoSvPmTF0nnpOG5De/2tN68KOuowAEcwjF4cAZNuIYWtAHDAzzBC7xaj9az9Wa9/4wuWeXOPvyB9fENyBqUTg==</latexit>GNdec | 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 | 788 | <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 | 789 | sha1_base64="79QHnx4t/4kSfeuqQfRiz+zDMfI=">AAAB+nicbVDLSsNAFJ3UV62vVJduBovgqiQi6MJFwYWupIJ9QBvCZDJph85MwsxEKTGf4saFIm79Enf+jZM2C209MHA4517umRMkjCrtON9WZWV1bX2julnb2t7Z3bPr+10VpxKTDo5ZLPsBUoRRQTqaakb6iSSIB4z0gslV4fceiFQ0Fvd6mhCPo5GgEcVIG8m360OO9Fjy7Po297OQ4Ny3G07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MS | 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 | 790 | 07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MSU0xI3ltmCqSIDxBIzIwVCBOlJfNoufw2CghjGJpntBwpv7eyBBXasoDM1kEVYteIf7nDVIdXXgZFUmqicDzQ1HKoI5h0QMMqSRYs6khCEtqskI8RhJhbdqqmRLcxS8vk+5p03Wa7t1Zo3VZ1lEFh+AInAAXnIMWuAFt0AEYPIJn8ArerCfrxXq3PuajFavcOQB/YH3+ALjdlEQ=</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 | 791 | sha1_base64="79QHnx4t/4kSfeuqQfRiz+zDMfI=">AAAB+nicbVDLSsNAFJ3UV62vVJduBovgqiQi6MJFwYWupIJ9QBvCZDJph85MwsxEKTGf4saFIm79Enf+jZM2C209MHA4517umRMkjCrtON9WZWV1bX2julnb2t7Z3bPr+10VpxKTDo5ZLPsBUoRRQTqaakb6iSSIB4z0gslV4fceiFQ0Fvd6mhCPo5GgEcVIG8m360OO9Fjy7Po297OQ4Ny3G07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MS | 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 | 792 | 07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MSU0xI3ltmCqSIDxBIzIwVCBOlJfNoufw2CghjGJpntBwpv7eyBBXasoDM1kEVYteIf7nDVIdXXgZFUmqicDzQ1HKoI5h0QMMqSRYs6khCEtqskI8RhJhbdqqmRLcxS8vk+5p03Wa7t1Zo3VZ1lEFh+AInAAXnIMWuAFt0AEYPIJn8ArerCfrxXq3PuajFavcOQB/YH3+ALjdlEQ=</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 | 793 | sha1_base64="79QHnx4t/4kSfeuqQfRiz+zDMfI=">AAAB+nicbVDLSsNAFJ3UV62vVJduBovgqiQi6MJFwYWupIJ9QBvCZDJph85MwsxEKTGf4saFIm79Enf+jZM2C209MHA4517umRMkjCrtON9WZWV1bX2julnb2t7Z3bPr+10VpxKTDo5ZLPsBUoRRQTqaakb6iSSIB4z0gslV4fceiFQ0Fvd6mhCPo5GgEcVIG8m360OO9Fjy7Po297OQ4Ny3G07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MS | 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 | 794 | 07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MSU0xI3ltmCqSIDxBIzIwVCBOlJfNoufw2CghjGJpntBwpv7eyBBXasoDM1kEVYteIf7nDVIdXXgZFUmqicDzQ1HKoI5h0QMMqSRYs6khCEtqskI8RhJhbdqqmRLcxS8vk+5p03Wa7t1Zo3VZ1lEFh+AInAAXnIMWuAFt0AEYPIJn8ArerCfrxXq3PuajFavcOQB/YH3+ALjdlEQ=</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 | 795 | sha1_base64="79QHnx4t/4kSfeuqQfRiz+zDMfI=">AAAB+nicbVDLSsNAFJ3UV62vVJduBovgqiQi6MJFwYWupIJ9QBvCZDJph85MwsxEKTGf4saFIm79Enf+jZM2C209MHA4517umRMkjCrtON9WZWV1bX2julnb2t7Z3bPr+10VpxKTDo5ZLPsBUoRRQTqaakb6iSSIB4z0gslV4fceiFQ0Fvd6mhCPo5GgEcVIG8m360OO9Fjy7Po297OQ4Ny3G07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MS | 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 | 796 | 07TmQEuE7ckDVCi7dtfwzDGKSdCY4aUGrhOor0MSU0xI3ltmCqSIDxBIzIwVCBOlJfNoufw2CghjGJpntBwpv7eyBBXasoDM1kEVYteIf7nDVIdXXgZFUmqicDzQ1HKoI5h0QMMqSRYs6khCEtqskI8RhJhbdqqmRLcxS8vk+5p03Wa7t1Zo3VZ1lEFh+AInAAXnIMWuAFt0AEYPIJn8ArerCfrxXq3PuajFavcOQB/YH3+ALjdlEQ=</latexit>GNcore | 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 | 797 | <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 | 798 | sha1_base64="sfcetjjriA53KVhP8LRkSGs9KNA=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNCVVLAPaEOYTCft0HmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0oY1cbzvp2V1bX1jc3KVnV7Z3dv3z2odbRMFSZtLJlUvQhpwqggbUMNI71EEcQjRrrR5Lrwu49EaSrFg5kmJOBoJGhMMTJWCt3agCMzVjy7ucvDDEtF8tCtew1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkD | 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 | 799 | w1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkDMWM5NVBqkmC8ASNSN9SgTjRQTbLnsMTqwxhLJV9wsCZ+nsjQ1zrKY/sZJFUL3qF+J/XT018GWRUJKkhAs8PxSmDRsKiCDikimDDppYgrKjNCvEYKYSNratqS/AXv7xMOmcN32v49+f15lVZRwUcgWNwCnxwAZrgFrRAG2DwBJ7BK3hzcufFeXc+5qMrTrlzCP7A+fwBopiUyw==</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 | 800 | sha1_base64="sfcetjjriA53KVhP8LRkSGs9KNA=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNCVVLAPaEOYTCft0HmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0oY1cbzvp2V1bX1jc3KVnV7Z3dv3z2odbRMFSZtLJlUvQhpwqggbUMNI71EEcQjRrrR5Lrwu49EaSrFg5kmJOBoJGhMMTJWCt3agCMzVjy7ucvDDEtF8tCtew1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkD | 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 | 801 | w1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkDMWM5NVBqkmC8ASNSN9SgTjRQTbLnsMTqwxhLJV9wsCZ+nsjQ1zrKY/sZJFUL3qF+J/XT018GWRUJKkhAs8PxSmDRsKiCDikimDDppYgrKjNCvEYKYSNratqS/AXv7xMOmcN32v49+f15lVZRwUcgWNwCnxwAZrgFrRAG2DwBJ7BK3hzcufFeXc+5qMrTrlzCP7A+fwBopiUyw==</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 | 802 | sha1_base64="sfcetjjriA53KVhP8LRkSGs9KNA=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNCVVLAPaEOYTCft0HmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0oY1cbzvp2V1bX1jc3KVnV7Z3dv3z2odbRMFSZtLJlUvQhpwqggbUMNI71EEcQjRrrR5Lrwu49EaSrFg5kmJOBoJGhMMTJWCt3agCMzVjy7ucvDDEtF8tCtew1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkD | 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 | 803 | w1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkDMWM5NVBqkmC8ASNSN9SgTjRQTbLnsMTqwxhLJV9wsCZ+nsjQ1zrKY/sZJFUL3qF+J/XT018GWRUJKkhAs8PxSmDRsKiCDikimDDppYgrKjNCvEYKYSNratqS/AXv7xMOmcN32v49+f15lVZRwUcgWNwCnxwAZrgFrRAG2DwBJ7BK3hzcufFeXc+5qMrTrlzCP7A+fwBopiUyw==</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 | 804 | sha1_base64="sfcetjjriA53KVhP8LRkSGs9KNA=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNCVVLAPaEOYTCft0HmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0oY1cbzvp2V1bX1jc3KVnV7Z3dv3z2odbRMFSZtLJlUvQhpwqggbUMNI71EEcQjRrrR5Lrwu49EaSrFg5kmJOBoJGhMMTJWCt3agCMzVjy7ucvDDEtF8tCtew1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkD | 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 | 805 | w1vBrhM/JLUQYlW6H4NhhKnnAiDGdK673uJCTKkDMWM5NVBqkmC8ASNSN9SgTjRQTbLnsMTqwxhLJV9wsCZ+nsjQ1zrKY/sZJFUL3qF+J/XT018GWRUJKkhAs8PxSmDRsKiCDikimDDppYgrKjNCvEYKYSNratqS/AXv7xMOmcN32v49+f15lVZRwUcgWNwCnxwAZrgFrRAG2DwBJ7BK3hzcufFeXc+5qMrTrlzCP7A+fwBopiUyw==</latexit>⇥M | 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 | 806 | <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 | 807 | sha1_base64="xCEPSgjeJaAOppNxwTZXrwRukIg=">AAAB73icbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYGMjRDAfkBxhb7OXLNnbu+zOCeHIn7CxUMTWv2Pnv3GTXKGJDwYe780wMy9IpDDout9OYW19Y3OruF3a2d3bPygfHjVNnGrGGyyWsW4H1HApFG+gQMnbieY0CiRvBaPbmd964tqIWD3iJOF+RAdKhIJRtFK7iyLihtz3yhW36s5BVomXkwrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsan | 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 | 808 | wrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsanlA2ogPesVRRu8bP5vdOyZlV+iSMtS2FZK7+nshoZMwkCmxnRHFolr2Z+J/XSTG89jOhkhS5YotFYSoJxmT2POkLzRnKiSWUaWFvJWxINWVoIyrZELzll1dJ86LquVXv4bJSu8njKMIJnMI5eHAFNbiDOjSAgYRneIU3Z+y8OO/Ox6K14OQzx/AHzucPqJqPrw==</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 | 809 | sha1_base64="xCEPSgjeJaAOppNxwTZXrwRukIg=">AAAB73icbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYGMjRDAfkBxhb7OXLNnbu+zOCeHIn7CxUMTWv2Pnv3GTXKGJDwYe780wMy9IpDDout9OYW19Y3OruF3a2d3bPygfHjVNnGrGGyyWsW4H1HApFG+gQMnbieY0CiRvBaPbmd964tqIWD3iJOF+RAdKhIJRtFK7iyLihtz3yhW36s5BVomXkwrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsan | 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 | 810 | wrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsanlA2ogPesVRRu8bP5vdOyZlV+iSMtS2FZK7+nshoZMwkCmxnRHFolr2Z+J/XSTG89jOhkhS5YotFYSoJxmT2POkLzRnKiSWUaWFvJWxINWVoIyrZELzll1dJ86LquVXv4bJSu8njKMIJnMI5eHAFNbiDOjSAgYRneIU3Z+y8OO/Ox6K14OQzx/AHzucPqJqPrw==</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 | 811 | sha1_base64="xCEPSgjeJaAOppNxwTZXrwRukIg=">AAAB73icbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYGMjRDAfkBxhb7OXLNnbu+zOCeHIn7CxUMTWv2Pnv3GTXKGJDwYe780wMy9IpDDout9OYW19Y3OruF3a2d3bPygfHjVNnGrGGyyWsW4H1HApFG+gQMnbieY0CiRvBaPbmd964tqIWD3iJOF+RAdKhIJRtFK7iyLihtz3yhW36s5BVomXkwrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsan | 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 | 812 | wrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsanlA2ogPesVRRu8bP5vdOyZlV+iSMtS2FZK7+nshoZMwkCmxnRHFolr2Z+J/XSTG89jOhkhS5YotFYSoJxmT2POkLzRnKiSWUaWFvJWxINWVoIyrZELzll1dJ86LquVXv4bJSu8njKMIJnMI5eHAFNbiDOjSAgYRneIU3Z+y8OO/Ox6K14OQzx/AHzucPqJqPrw==</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 | 813 | sha1_base64="xCEPSgjeJaAOppNxwTZXrwRukIg=">AAAB73icbVA9SwNBEJ2LXzF+RS1tFoNgFe5E0MIiYGMjRDAfkBxhb7OXLNnbu+zOCeHIn7CxUMTWv2Pnv3GTXKGJDwYe780wMy9IpDDout9OYW19Y3OruF3a2d3bPygfHjVNnGrGGyyWsW4H1HApFG+gQMnbieY0CiRvBaPbmd964tqIWD3iJOF+RAdKhIJRtFK7iyLihtz3yhW36s5BVomXkwrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsan | 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 | 814 | wrkqPfKX91+zNKIK2SSGtPx3AT9jGoUTPJpqZsanlA2ogPesVRRu8bP5vdOyZlV+iSMtS2FZK7+nshoZMwkCmxnRHFolr2Z+J/XSTG89jOhkhS5YotFYSoJxmT2POkLzRnKiSWUaWFvJWxINWVoIyrZELzll1dJ86LquVXv4bJSu8njKMIJnMI5eHAFNbiDOjSAgYRneIU3Z+y8OO/Ox6K14OQzx/AHzucPqJqPrw==</latexit>Gthid | 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 | 815 | <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 | 816 | sha1_base64="Vr1kZJh7jApVU4cGmtPimK4PPic=">AAAB/XicbVDLSsNAFJ3UV62v+Ni5GSyCq5KIoAsXBRe6rGAf0MYwmUzaoTOTMDMRagj+ihsXirj1P9z5N07aLLT1wMDhnHu5Z06QMKq043xblaXlldW16nptY3Nre8fe3euoOJWYtHHMYtkLkCKMCtLWVDPSSyRBPGCkG4yvCr/7QKSisbjTk4R4HA0FjShG2ki+fXB9n+nczwYc6ZHk2YiGee7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0 | 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 | 817 | e7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0V6GpKaYkbw2SBVJEB6jIekbKhAnysum6XN4bJQQRrE0T2g4VX9vZIgrNeGBmSxCqnmvEP/z+qmOLryMiiTVRODZoShlUMewqAKGVBKs2cQQhCU1WSEeIYmwNoXVTAnu/JcXSee04ToN9/as3rws66iCQ3AEToALzkET3IAWaAMMHsEzeAVv1pP1Yr1bH7PRilXu7IM/sD5/AI1Vlec=</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 | 818 | sha1_base64="Vr1kZJh7jApVU4cGmtPimK4PPic=">AAAB/XicbVDLSsNAFJ3UV62v+Ni5GSyCq5KIoAsXBRe6rGAf0MYwmUzaoTOTMDMRagj+ihsXirj1P9z5N07aLLT1wMDhnHu5Z06QMKq043xblaXlldW16nptY3Nre8fe3euoOJWYtHHMYtkLkCKMCtLWVDPSSyRBPGCkG4yvCr/7QKSisbjTk4R4HA0FjShG2ki+fXB9n+nczwYc6ZHk2YiGee7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0 | 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 | 819 | e7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0V6GpKaYkbw2SBVJEB6jIekbKhAnysum6XN4bJQQRrE0T2g4VX9vZIgrNeGBmSxCqnmvEP/z+qmOLryMiiTVRODZoShlUMewqAKGVBKs2cQQhCU1WSEeIYmwNoXVTAnu/JcXSee04ToN9/as3rws66iCQ3AEToALzkET3IAWaAMMHsEzeAVv1pP1Yr1bH7PRilXu7IM/sD5/AI1Vlec=</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 | 820 | sha1_base64="Vr1kZJh7jApVU4cGmtPimK4PPic=">AAAB/XicbVDLSsNAFJ3UV62v+Ni5GSyCq5KIoAsXBRe6rGAf0MYwmUzaoTOTMDMRagj+ihsXirj1P9z5N07aLLT1wMDhnHu5Z06QMKq043xblaXlldW16nptY3Nre8fe3euoOJWYtHHMYtkLkCKMCtLWVDPSSyRBPGCkG4yvCr/7QKSisbjTk4R4HA0FjShG2ki+fXB9n+nczwYc6ZHk2YiGee7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0 | 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 | 821 | e7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0V6GpKaYkbw2SBVJEB6jIekbKhAnysum6XN4bJQQRrE0T2g4VX9vZIgrNeGBmSxCqnmvEP/z+qmOLryMiiTVRODZoShlUMewqAKGVBKs2cQQhCU1WSEeIYmwNoXVTAnu/JcXSee04ToN9/as3rws66iCQ3AEToALzkET3IAWaAMMHsEzeAVv1pP1Yr1bH7PRilXu7IM/sD5/AI1Vlec=</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 | 822 | sha1_base64="Vr1kZJh7jApVU4cGmtPimK4PPic=">AAAB/XicbVDLSsNAFJ3UV62v+Ni5GSyCq5KIoAsXBRe6rGAf0MYwmUzaoTOTMDMRagj+ihsXirj1P9z5N07aLLT1wMDhnHu5Z06QMKq043xblaXlldW16nptY3Nre8fe3euoOJWYtHHMYtkLkCKMCtLWVDPSSyRBPGCkG4yvCr/7QKSisbjTk4R4HA0FjShG2ki+fXB9n+nczwYc6ZHk2YiGee7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0 | 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 | 823 | e7bdafhTAEXiVuSOijR8u2vQRjjlBOhMUNK9V0n0V6GpKaYkbw2SBVJEB6jIekbKhAnysum6XN4bJQQRrE0T2g4VX9vZIgrNeGBmSxCqnmvEP/z+qmOLryMiiTVRODZoShlUMewqAKGVBKs2cQQhCU1WSEeIYmwNoXVTAnu/JcXSee04ToN9/as3rws66iCQ3AEToALzkET3IAWaAMMHsEzeAVv1pP1Yr1bH7PRilXu7IM/sD5/AI1Vlec=</latexit>Gt 1hid | 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 | 824 | <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 | 825 | sha1_base64="TTYaLKTJF7FnWkmBKBRepTsjYHc=">AAAB/3icbVDLSsNAFJ34rPUVFdy4GSyCG0sigi5cFFzosoJ9QBvDZDJph85MwsxEKDELf8WNC0Xc+hvu/BsnbRbaemDgcM693DMnSBhV2nG+rYXFpeWV1cpadX1jc2vb3tltqziVmLRwzGLZDZAijArS0lQz0k0kQTxgpBOMrgq/80CkorG40+OEeBwNBI0oRtpIvr1/fZ/pEzf3sz5Heih5NqRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSP | 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 | 826 | qRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSPddJtJchqSlmJK/2U0UShEdoQHqGCsSJ8rJJ/hweGSWEUSzNExpO1N8bGeJKjXlgJouQatYrxP+8XqqjCy+jIkk1EXh6KEoZ1DEsyoAhlQRrNjYEYUlNVoiHSCKsTWVVU4I7++V50j6tu07dvT2rNS7LOirgAByCY+CCc9AAN6AJWgCDR/AMXsGb9WS9WO/Wx3R0wSp39sAfWJ8/b2WWWQ==</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 | 827 | sha1_base64="TTYaLKTJF7FnWkmBKBRepTsjYHc=">AAAB/3icbVDLSsNAFJ34rPUVFdy4GSyCG0sigi5cFFzosoJ9QBvDZDJph85MwsxEKDELf8WNC0Xc+hvu/BsnbRbaemDgcM693DMnSBhV2nG+rYXFpeWV1cpadX1jc2vb3tltqziVmLRwzGLZDZAijArS0lQz0k0kQTxgpBOMrgq/80CkorG40+OEeBwNBI0oRtpIvr1/fZ/pEzf3sz5Heih5NqRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSP | 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 | 828 | qRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSPddJtJchqSlmJK/2U0UShEdoQHqGCsSJ8rJJ/hweGSWEUSzNExpO1N8bGeJKjXlgJouQatYrxP+8XqqjCy+jIkk1EXh6KEoZ1DEsyoAhlQRrNjYEYUlNVoiHSCKsTWVVU4I7++V50j6tu07dvT2rNS7LOirgAByCY+CCc9AAN6AJWgCDR/AMXsGb9WS9WO/Wx3R0wSp39sAfWJ8/b2WWWQ==</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 | 829 | sha1_base64="TTYaLKTJF7FnWkmBKBRepTsjYHc=">AAAB/3icbVDLSsNAFJ34rPUVFdy4GSyCG0sigi5cFFzosoJ9QBvDZDJph85MwsxEKDELf8WNC0Xc+hvu/BsnbRbaemDgcM693DMnSBhV2nG+rYXFpeWV1cpadX1jc2vb3tltqziVmLRwzGLZDZAijArS0lQz0k0kQTxgpBOMrgq/80CkorG40+OEeBwNBI0oRtpIvr1/fZ/pEzf3sz5Heih5NqRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSP | 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 | 830 | qRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSPddJtJchqSlmJK/2U0UShEdoQHqGCsSJ8rJJ/hweGSWEUSzNExpO1N8bGeJKjXlgJouQatYrxP+8XqqjCy+jIkk1EXh6KEoZ1DEsyoAhlQRrNjYEYUlNVoiHSCKsTWVVU4I7++V50j6tu07dvT2rNS7LOirgAByCY+CCc9AAN6AJWgCDR/AMXsGb9WS9WO/Wx3R0wSp39sAfWJ8/b2WWWQ==</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 | 831 | sha1_base64="TTYaLKTJF7FnWkmBKBRepTsjYHc=">AAAB/3icbVDLSsNAFJ34rPUVFdy4GSyCG0sigi5cFFzosoJ9QBvDZDJph85MwsxEKDELf8WNC0Xc+hvu/BsnbRbaemDgcM693DMnSBhV2nG+rYXFpeWV1cpadX1jc2vb3tltqziVmLRwzGLZDZAijArS0lQz0k0kQTxgpBOMrgq/80CkorG40+OEeBwNBI0oRtpIvr1/fZ/pEzf3sz5Heih5NqRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSP | 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 | 832 | qRhnvt2zak7E8B54pakBko0ffurH8Y45URozJBSPddJtJchqSlmJK/2U0UShEdoQHqGCsSJ8rJJ/hweGSWEUSzNExpO1N8bGeJKjXlgJouQatYrxP+8XqqjCy+jIkk1EXh6KEoZ1DEsyoAhlQRrNjYEYUlNVoiHSCKsTWVVU4I7++V50j6tu07dvT2rNS7LOirgAByCY+CCc9AAN6AJWgCDR/AMXsGb9WS9WO/Wx3R0wSp39sAfWJ8/b2WWWQ==</latexit>Gtout | 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 | 833 | <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 | 834 | sha1_base64="Fm4Iz64LvWn082kf2+Dq6sd1g9I=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNBlBfuANobJdNIOnWTCzI1YQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0gE1+A439bK6tr6xmZlq7q9s7u3bx/UOlqmirI2lUKqXkA0EzxmbeAgWC9RjESBYN1gcl343UemNJfxPUwT5kVkFPOQUwJG8u3azQP42SAiMFZRJlPIc9+uOw1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKig | 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 | 835 | w1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKigFPB8uog1SwhdEJGrG9oTCKmvWyWPccnRhniUCrzYsAz9fdGRiKtp1FgJouQetErxP+8fgrhpZfxOEmBxXR+KEwFBomLIvCQK0ZBTA0hVHGTFdMxUYSCqatqSnAXv7xMOmcN12m4d+f15lVZRwUdoWN0ilx0gZroFrVQG1H0hJ7RK3qzcuvFerc+5qMrVrlziP7A+vwB7zuU/g==</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 | 836 | sha1_base64="Fm4Iz64LvWn082kf2+Dq6sd1g9I=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNBlBfuANobJdNIOnWTCzI1YQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0gE1+A439bK6tr6xmZlq7q9s7u3bx/UOlqmirI2lUKqXkA0EzxmbeAgWC9RjESBYN1gcl343UemNJfxPUwT5kVkFPOQUwJG8u3azQP42SAiMFZRJlPIc9+uOw1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKig | 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 | 837 | w1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKigFPB8uog1SwhdEJGrG9oTCKmvWyWPccnRhniUCrzYsAz9fdGRiKtp1FgJouQetErxP+8fgrhpZfxOEmBxXR+KEwFBomLIvCQK0ZBTA0hVHGTFdMxUYSCqatqSnAXv7xMOmcN12m4d+f15lVZRwUdoWN0ilx0gZroFrVQG1H0hJ7RK3qzcuvFerc+5qMrVrlziP7A+vwB7zuU/g==</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 | 838 | sha1_base64="Fm4Iz64LvWn082kf2+Dq6sd1g9I=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNBlBfuANobJdNIOnWTCzI1YQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0gE1+A439bK6tr6xmZlq7q9s7u3bx/UOlqmirI2lUKqXkA0EzxmbeAgWC9RjESBYN1gcl343UemNJfxPUwT5kVkFPOQUwJG8u3azQP42SAiMFZRJlPIc9+uOw1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKig | 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 | 839 | w1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKigFPB8uog1SwhdEJGrG9oTCKmvWyWPccnRhniUCrzYsAz9fdGRiKtp1FgJouQetErxP+8fgrhpZfxOEmBxXR+KEwFBomLIvCQK0ZBTA0hVHGTFdMxUYSCqatqSnAXv7xMOmcN12m4d+f15lVZRwUdoWN0ilx0gZroFrVQG1H0hJ7RK3qzcuvFerc+5qMrVrlziP7A+vwB7zuU/g==</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 | 840 | sha1_base64="Fm4Iz64LvWn082kf2+Dq6sd1g9I=">AAAB+3icbVDLSsNAFJ34rPUV69LNYBFclUQEXbgouNBlBfuANobJdNIOnWTCzI1YQn7FjQtF3Poj7vwbJ20W2npg4HDOvdwzJ0gE1+A439bK6tr6xmZlq7q9s7u3bx/UOlqmirI2lUKqXkA0EzxmbeAgWC9RjESBYN1gcl343UemNJfxPUwT5kVkFPOQUwJG8u3azQP42SAiMFZRJlPIc9+uOw1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKig | 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 | 841 | w1nBrxM3JLUUYmWb38NhpKmEYuBCqJ133US8DKigFPB8uog1SwhdEJGrG9oTCKmvWyWPccnRhniUCrzYsAz9fdGRiKtp1FgJouQetErxP+8fgrhpZfxOEmBxXR+KEwFBomLIvCQK0ZBTA0hVHGTFdMxUYSCqatqSnAXv7xMOmcN12m4d+f15lVZRwUdoWN0ilx0gZroFrVQG1H0hJ7RK3qzcuvFerc+5qMrVrlziP7A+vwB7zuU/g==</latexit>Gtinp | 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 | 842 | <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 | 843 | sha1_base64="KNSYXpe+JERsrbP/+QxnbJ2F1T0=">AAAB+3icbVDLSsNAFL3xWesr1qWbYBFclUQEXbgouNBlBfuANobJdNIOnZmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvcy5J0wYVdp1v62V1bX1jc3KVnV7Z3dv3z6odVScSkzaOGax7IVIEUYFaWuqGeklkiAeMtINJ9eF330kUtFY3OtpQnyORoJGFCNtpMCu3QTZgCM9ljyjIsnzBx3YdbfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MS | 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 | 844 | bfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MSU0xI3l1kCqSIDxBI9I3VCBOlJ/NsufOiVGGThRL84R2ZurvjQxxpaY8NJNFTLXoFeJ/Xj/V0aVf3JRqIvD8oyhljo6doghnSCXBmk0NQVhSk9XBYyQR1qauqinBWzx5mXTOGp7b8O7O682rso4KHMExnIIHF9CEW2hBGzA8wTO8wpuVWy/Wu/UxH12xyp1D+APr8wfV6pTt</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 | 845 | sha1_base64="KNSYXpe+JERsrbP/+QxnbJ2F1T0=">AAAB+3icbVDLSsNAFL3xWesr1qWbYBFclUQEXbgouNBlBfuANobJdNIOnZmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvcy5J0wYVdp1v62V1bX1jc3KVnV7Z3dv3z6odVScSkzaOGax7IVIEUYFaWuqGeklkiAeMtINJ9eF330kUtFY3OtpQnyORoJGFCNtpMCu3QTZgCM9ljyjIsnzBx3YdbfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MS | 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 | 846 | bfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MSU0xI3l1kCqSIDxBI9I3VCBOlJ/NsufOiVGGThRL84R2ZurvjQxxpaY8NJNFTLXoFeJ/Xj/V0aVf3JRqIvD8oyhljo6doghnSCXBmk0NQVhSk9XBYyQR1qauqinBWzx5mXTOGp7b8O7O682rso4KHMExnIIHF9CEW2hBGzA8wTO8wpuVWy/Wu/UxH12xyp1D+APr8wfV6pTt</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 | 847 | sha1_base64="KNSYXpe+JERsrbP/+QxnbJ2F1T0=">AAAB+3icbVDLSsNAFL3xWesr1qWbYBFclUQEXbgouNBlBfuANobJdNIOnZmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvcy5J0wYVdp1v62V1bX1jc3KVnV7Z3dv3z6odVScSkzaOGax7IVIEUYFaWuqGeklkiAeMtINJ9eF330kUtFY3OtpQnyORoJGFCNtpMCu3QTZgCM9ljyjIsnzBx3YdbfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MS | 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 | 848 | bfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MSU0xI3l1kCqSIDxBI9I3VCBOlJ/NsufOiVGGThRL84R2ZurvjQxxpaY8NJNFTLXoFeJ/Xj/V0aVf3JRqIvD8oyhljo6doghnSCXBmk0NQVhSk9XBYyQR1qauqinBWzx5mXTOGp7b8O7O682rso4KHMExnIIHF9CEW2hBGzA8wTO8wpuVWy/Wu/UxH12xyp1D+APr8wfV6pTt</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 | 849 | sha1_base64="KNSYXpe+JERsrbP/+QxnbJ2F1T0=">AAAB+3icbVDLSsNAFL3xWesr1qWbYBFclUQEXbgouNBlBfuANobJdNIOnZmEmYlYQn7FjQtF3Poj7vwbJ20W2npg4HDOvcy5J0wYVdp1v62V1bX1jc3KVnV7Z3dv3z6odVScSkzaOGax7IVIEUYFaWuqGeklkiAeMtINJ9eF330kUtFY3OtpQnyORoJGFCNtpMCu3QTZgCM9ljyjIsnzBx3YdbfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MS | 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 | 850 | bfhzuAsE68kdSjRCuyvwTDGKSdCY4aU6ntuov0MSU0xI3l1kCqSIDxBI9I3VCBOlJ/NsufOiVGGThRL84R2ZurvjQxxpaY8NJNFTLXoFeJ/Xj/V0aVf3JRqIvD8oyhljo6doghnSCXBmk0NQVhSk9XBYyQR1qauqinBWzx5mXTOGp7b8O7O682rso4KHMExnIIHF9CEW2hBGzA8wTO8wpuVWy/Wu/UxH12xyp1D+APr8wfV6pTt</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 | 851 | (c) Recurrent GN architecture | 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 | 852 | Figure 6: (a) An example composing multiple GN blocks in sequence to form a GN \core". Here,
the GN blocks can use shared weights, or they could be independent. (b) The encode-process-decode
architecture, which is a common choice for composing GN blocks (see Section 4.3). Here, a GN
encodes an input graph, which is then processed by a GN core. The output of the core is decoded
by a third GN block into an output graph, whose nodes, edges, and/or global attributes would be
used for task-speci c purposes. (c) The encode-process-decode architecture applied in a sequential
setting in which the core is also unrolled over time (potentially using a GRU or LSTM architecture),
in addition to being repeated within each time step. Here, merged lines indicate concatenation, and
split lines indicate copying.
4.3 Composable multi-block architectures
A key design principle of graph networks is constructing complex architectures by composing GN
blocks. We de ned a GN block as always taking a graph comprised of edge, node, and global
elements as input, and returning a graph with the same constituent elements as output (simply
passing through the input elements to the output when those elements are not explicitly updated). | 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 | 853 | elements as input, and returning a graph with the same constituent elements as output (simply
passing through the input elements to the output when those elements are not explicitly updated).
This graph-to-graph input/output interface ensures that the output of one GN block can be passed
as input to another, even if their internal con gurations are di erent, similar to the tensor-to-tensor
interface of the standard deep learning toolkit. In the most basic form, two GN blocks, GN1and
GN2, can be composed as GN1GN2by passing the output of the rst as input to the second:
G0= GN 2(GN 1(G)).
Arbitrary numbers of GN blocks can be composed, as show in Figure 6a. The blocks can
be unshared (di erent functions and/or parameters, analogous to layers of a CNN), GN16=
GN26=6=GNM, or shared (reused functions and parameters, analogous to an unrolled RNN),
GN1=GN2==GNM. The white box around the GNcorein Figure 6a represents Mrepeated
internal processing sub-steps, with either shared or unshared GN blocks. Shared con gurations
are analogous to message-passing (Gilmer et al., 2017), where the same local update procedure is | 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 | 854 | are analogous to message-passing (Gilmer et al., 2017), where the same local update procedure is
applied iteratively to propagate information across the structure (Figure 7). If we exclude the global
u(which aggregates information from across the nodes and edges), the information that a node
has access to after msteps of propagation is determined by the set of nodes and edges that are at
mostmhops away. This can be interpreted as breaking down a complex computation into smaller
elementary steps. The steps can also be used to capture sequentiality in time. In our ball-spring
example, if each propagation step predicts the physical dynamics over one time step of duration t,
then theMpropagation steps result in a total simulation time of, Mt.
A common architecture design is what we call the encode-process-decode con guration (Hamrick
et al. (2018); also see Figure 6ba): an input graph, Ginpis transformed into a latent representation,
G0, by an encoder, GNenc; a shared core block, GNcore, is applied Mtimes to return GM; and
nally an output graph, Gout, is decoded by GNdec. For example, in our running example, the
encoder might compute the initial forces and interaction energies between the balls, the core might
19
m=0 | 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 | 855 | <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 | 856 | sha1_base64="ic6WezPV7TWar890N1QnpVOPjNA=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoOQU9gVQS9CwIvHiOYByRJmJ7PJkJnZZaZXCCGf4MWDIl79Im/+jZNkDxotaCiquunuilIpLPr+l1dYW9/Y3Cpul3Z29/YPyodHLZtkhvEmS2RiOhG1XArNmyhQ8k5qOFWR5O1ofDP324/cWJHoB5ykPFR0qEUsGEUn3atrv1+u+DV/AfKXBDmpQI5Gv/zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65 | 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 | 857 | /zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65F1HNVXchtPFqTNy5pQBiRPjSiNZqD8nplRZO1GR61QUR3bVm4v/ed0M46twKnSaIddsuSjOJMGEzP8mA2E4QzlxhDIj3K2EjaihDF06JRdCsPryX9I6rwV+Lbi7qNSreRxFOIFTqEIAl1CHW2hAExgM4Qle4NWT3rP35r0vWwtePnMMv+B9fAO9341Y</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 | 858 | sha1_base64="ic6WezPV7TWar890N1QnpVOPjNA=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoOQU9gVQS9CwIvHiOYByRJmJ7PJkJnZZaZXCCGf4MWDIl79Im/+jZNkDxotaCiquunuilIpLPr+l1dYW9/Y3Cpul3Z29/YPyodHLZtkhvEmS2RiOhG1XArNmyhQ8k5qOFWR5O1ofDP324/cWJHoB5ykPFR0qEUsGEUn3atrv1+u+DV/AfKXBDmpQI5Gv/zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65 | 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 | 859 | /zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65F1HNVXchtPFqTNy5pQBiRPjSiNZqD8nplRZO1GR61QUR3bVm4v/ed0M46twKnSaIddsuSjOJMGEzP8mA2E4QzlxhDIj3K2EjaihDF06JRdCsPryX9I6rwV+Lbi7qNSreRxFOIFTqEIAl1CHW2hAExgM4Qle4NWT3rP35r0vWwtePnMMv+B9fAO9341Y</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 | 860 | sha1_base64="ic6WezPV7TWar890N1QnpVOPjNA=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoOQU9gVQS9CwIvHiOYByRJmJ7PJkJnZZaZXCCGf4MWDIl79Im/+jZNkDxotaCiquunuilIpLPr+l1dYW9/Y3Cpul3Z29/YPyodHLZtkhvEmS2RiOhG1XArNmyhQ8k5qOFWR5O1ofDP324/cWJHoB5ykPFR0qEUsGEUn3atrv1+u+DV/AfKXBDmpQI5Gv/zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65 | 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 | 861 | /zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65F1HNVXchtPFqTNy5pQBiRPjSiNZqD8nplRZO1GR61QUR3bVm4v/ed0M46twKnSaIddsuSjOJMGEzP8mA2E4QzlxhDIj3K2EjaihDF06JRdCsPryX9I6rwV+Lbi7qNSreRxFOIFTqEIAl1CHW2hAExgM4Qle4NWT3rP35r0vWwtePnMMv+B9fAO9341Y</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 | 862 | sha1_base64="ic6WezPV7TWar890N1QnpVOPjNA=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoOQU9gVQS9CwIvHiOYByRJmJ7PJkJnZZaZXCCGf4MWDIl79Im/+jZNkDxotaCiquunuilIpLPr+l1dYW9/Y3Cpul3Z29/YPyodHLZtkhvEmS2RiOhG1XArNmyhQ8k5qOFWR5O1ofDP324/cWJHoB5ykPFR0qEUsGEUn3atrv1+u+DV/AfKXBDmpQI5Gv/zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65 | 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 | 863 | /zZGyQsU1wjk9TabuCnGE6pQcEkn5V6meUpZWM65F1HNVXchtPFqTNy5pQBiRPjSiNZqD8nplRZO1GR61QUR3bVm4v/ed0M46twKnSaIddsuSjOJMGEzP8mA2E4QzlxhDIj3K2EjaihDF06JRdCsPryX9I6rwV+Lbi7qNSreRxFOIFTqEIAl1CHW2hAExgM4Qle4NWT3rP35r0vWwtePnMMv+B9fAO9341Y</latexit>m=1 | 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 | 864 | <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 | 865 | sha1_base64="Zv5ybfKo/H8QvXQaOWqQVyOFtg4=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoOQU9gVQS9CwIvHiOYByRJmJ73JkJnZZWZWCCGf4MWDIl79Im/+jZNkDxotaCiquunuilLBjfX9L6+wtr6xuVXcLu3s7u0flA+PWibJNMMmS0SiOxE1KLjCpuVWYCfVSGUksB2Nb+Z++xG14Yl6sJMUQ0mHisecUeuke3kd9MsVv+YvQP6SICcVyNHolz97g4RlEpVlghrTDfzUhlOqLWcCZ6VeZjClbEyH2 | 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 | 866 | z97g4RlEpVlghrTDfzUhlOqLWcCZ6VeZjClbEyH2HVUUYkmnC5OnZEzpwxInGhXypKF+nNiSqUxExm5TkntyKx6c/E/r5vZ+CqccpVmFhVbLoozQWxC5n+TAdfIrJg4Qpnm7lbCRlRTZl06JRdCsPryX9I6rwV+Lbi7qNSreRxFOIFTqEIAl1CHW2hAExgM4Qle4NUT3rP35r0vWwtePnMMv+B9fAO/Y41Z</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 | 867 | sha1_base64="Zv5ybfKo/H8QvXQaOWqQVyOFtg4=">AAAB6nicbVDLSgNBEOyNrxhfUY9eBoOQU9gVQS9CwIvHiOYByRJmJ73JkJnZZWZWCCGf4MWDIl79Im/+jZNkDxotaCiquunuilLBjfX9L6+wtr6xuVXcLu3s7u0flA+PWibJNMMmS0SiOxE1KLjCpuVWYCfVSGUksB2Nb+Z++xG14Yl6sJMUQ0mHisecUeuke3kd9MsVv+YvQP6SICcVyNHolz97g4RlEpVlghrTDfzUhlOqLWcCZ6VeZjClbEyH2 | 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.