doi
stringlengths 10
10
| chunk-id
int64 0
936
| chunk
stringlengths 401
2.02k
| id
stringlengths 12
14
| title
stringlengths 8
162
| summary
stringlengths 228
1.92k
| source
stringlengths 31
31
| authors
stringlengths 7
6.97k
| categories
stringlengths 5
107
| comment
stringlengths 4
398
⌀ | journal_ref
stringlengths 8
194
⌀ | primary_category
stringlengths 5
17
| published
stringlengths 8
8
| updated
stringlengths 8
8
| references
list |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1701.07274 | 152 | Some recent papers follow: Asri et al. (2016), Bordes et al. (2017), Chen et al. (2016c), Eric and Manning (2017), Fatemi et al. (2016), Kandasamy et al. (2017), Lewis et al. (2017), Li et al. (2016a), Li et al. (2017a), Li et al. (2017b), Lipton et al. (2016), Mesnil et al. (2015), Mo et al. (2016), Peng et al. (2017a), Saon et al. (2016), Serban et al. (2017), Shah et al. (2016), She and Chai (2017), Su et al. (2016a), Weiss et al. (2017), Wen et al. (2015a), Wen et al. (2017), Williams and Zweig (2016), Williams et al. (2017), Xiong et al. (2017b), Xiong et al. (2017), Yang et al. (2016), Zhang et al. (2017a), Zhang et al. (2017c), Zhao and Eskenazi (2016), Zhou et al. (2017). See Serban et al. (2015) for a survey of corpora for building dialogue systems. | 1701.07274#152 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 153 | See NIPS 2016 Workshop on End-to-end Learning for Speech and Audio Processing, and NIPS 2015 Workshop on Machine Learning for Spoken Language Understanding and Interactions.
# 5.3.2 MACHINE TRANSLATION
Neural machine translation (Kalchbrenner and Blunsom, 2013; Cho et al., 2014; Sutskever et al., 2014; Bahdanau et al., 2015) utilizes end-to-end deep learning for machine translation, and becomes dominant, against the traditional statistical machine translation techniques. The neural machine translation approach usually ï¬rst encodes a variable-length source sentence, and then decodes it to a variable-length target sentence. Cho et al. (2014) and Sutskever et al. (2014) used two RNNs to encode a sentence to a ï¬x-length vector and then decode the vector into a target sentence. Bahdanau et al. (2015) introduced the soft-attention technique to learn to jointly align and translate. | 1701.07274#153 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 154 | He et al. (2016a) proposed dual learning mechanism to tackle the data hunger issue in machine translation, inspired by the observation that the information feedback between the primal, translation from language A to language B, and the dual, translation from B to A, can help improve both translation models, with a policy gradient method, using the language model likelihood as the reward signal. Experiments showed that, with only 10% bilingual data for warm start and monolingual data, the dual learning approach performed comparably with previous neural machine translation methods with full bilingual data in English to French tasks. The dual learning mechanism may have extensions to many tasks, if the task has a dual form, e.g., speech recognition and text to speech, image caption and image generation, question answering and question generation, search and keyword extraction, etc. | 1701.07274#154 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 155 | See Wu et al. (2016); Johnson et al. (2016) for Googleâs Neural Machine Translation System; Gehring et al. (2017) for convolutional sequence to sequence learning for fast neural machine trans- lation; Klein et al. (2017) for OpenNMT, an open source neural machine translation system; Cheng et al. (2016) for semi-supervised learning for neural machine translation, and Wu et al. (2017c) for adversarial neural machine translation. See Vaswani et al. (2017) for a new approach for translation that replaces CNN and RNN with attention and positional encoding. See Zhang et al. (2017b) for an open source toolkit for neural machine translation. See Monroe (2017) for a gentle introduction to translation.
Artetxe et al. (2017)
# 5.3.3 TEXT GENERATION
Text generation is the basis for many NLP problems, like conversational response generation, ma- chine translation, abstractive summarization, etc. | 1701.07274#155 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 156 | # 5.3.3 TEXT GENERATION
Text generation is the basis for many NLP problems, like conversational response generation, ma- chine translation, abstractive summarization, etc.
Text generation models are usually based on n-gram, feed-forward neural networks, or recurrent neural networks, trained to predict next word given the previous ground truth words as inputs; then in testing, the trained models are used to generate a sequence word by word, using the generated words as inputs. The errors will accumulate on the way, causing the exposure bias issue. Moreover, these models are trained with word level losses, e.g., cross entropy, to maximize the probability of next word; however, the models are evaluated on a different metrics like BLEU.
37
Ranzato et al. (2016) proposed Mixed Incremental Cross-Entropy Reinforce (MIXER) for sequence prediction, with incremental learning and a loss function combining both REINFORCE and cross- entropy. MIXER is a sequence level training algorithm, aligning training and testing objective, such as BLEU, rather than predicting the next word as in previous works. | 1701.07274#156 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 157 | Bahdanau et al. (2017) proposed an actor-critic algorithm for sequence prediction, attempting to further improve Ranzato et al. (2016). The authors utilized a critic network to predict the value of a token, i.e., the expected score following the sequence prediction policy, deï¬ned by an actor network, trained by the predicted value of tokens. Some techniques are deployed to improve performance: SARSA rather than Monter-Carlo method to lessen the variance in estimating value functions; target network for stability; sampling prediction from a delayed actor whose weights are updated more slowly than the actor to be trained, to avoid the feedback loop when actor and critic need to be trained based on the output of each other; reward shaping to avoid the issue of sparse training signal.
Yu et al. (2017) proposed SeqGAN, sequence generative adversarial nets with policy gradient, inte- grating the adversarial scheme in Goodfellow et al. (2014). Li et al. (2017a) proposed to improve sequence generation by considering the knowledge about the future.
5.4 COMPUTER VISION
Computer vision is about how computers gain understanding from digital images or videos. In the following, after presenting background in computer vision, we discuss recognition, motion analysis, scene understanding, integration with NLP, and visual control. | 1701.07274#157 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 158 | Reinforcement learning would be an important ingredient for interactive perception (Bohg et al., 2017), where perception and interaction with the environment would be helpful to each other, in tasks like object segmentation, articulation model estimation, object dynamics learning and haptic property estimation, object recognition or categorization, multimodal object model learning, object pose estimation, grasp planning, and manipulation skill learning.
More topics about applying deep RL to computer vision:
⢠Liu et al. (2017) for semantic parsing of large-scale 3D point clouds;
Devrim Kaba et al. (2017) for view planning, which is a set cover problem;
⢠Cao et al. (2017) for face hallucination, i.e., generating a high-resolution face image from a low-resolution input image;
Brunner et al. (2018) for learning to read maps;
⢠Bhatti et al. (2016) for SLAM-augmented DQN.
5.4.1 BACKGROUND | 1701.07274#158 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 159 | ⢠Bhatti et al. (2016) for SLAM-augmented DQN.
5.4.1 BACKGROUND
Todo: AlexNet (Krizhevsky et al., 2012), ResNets (He et al., 2016d) DenseNets (Huang et al., 2017), Fast R-CNN (Girshick, 2015), Faster R-CNN Ren et al. (2015), Mask R-CNN He et al. (2017), Shrivastava et al. (2017), VAEs (variational autoencoder) (Diederik P Kingma, 2014)
Todo: GANs (Goodfellow et al., 2014; Goodfellow, 2017); CycleGAN (Zhu et al., 2017a), Dual- GAN (Yi et al., 2017); See Arjovsky et al. (2017) for Wasserstein GAN (WGAN) as a stable GANs model. Gulrajani et al. (2017) proposed to improve stability of WGAN by penalizing the norm of the gradient of the discriminator with respect to its input, instead of clipping weights as in Arjovsky et al. (2017). Mao et al. (2016) proposed Least Squares GANs (LSGANs), another stable model. | 1701.07274#159 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 160 | Connection with RL: Finn et al. (2016a) established a connection between GANs, inverse RL, and energy-based models. Pfau and Vinyals (2016) established the connection between GANs and actor- critic algorithms. Ho and Ermon (2016) and Li et al. (2017) studied the connection between GANs and imitation learning.
autoencoder (Hinton and Salakhutdinov, 2006)
For disentangled factor learning, Kulkarni et al. (2015) proposed DC-IGN, the Deep Convolution Inverse Graphics Network, which follows a semi-supervised way; and Chen et al. (2016a) proposed InfoGAN, an information-theoretic extension to the Generative Adversarial Network, which follows
38
an unsupervised way. Zhou et al. (2015) showed that object detectors emerge from learning to recognize scenes, without supervised labels for objects.
Higgins et al. (2017) proposed β-VAE to automatically discover interpretable, disentangled, fac- torised, latent representations from raw images in an unsupervised way. The hyperparameter β balances latent channel capacity and independence constraints with reconstruction accuracy. When β = 1, β-VAE is the same as the original VAEs. | 1701.07274#160 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 161 | Eslami et al. (2016) proposed the framework of Attend-Infer-Repeat for efï¬cient inference in struc- tured image models to reason about objects explicitly. The authors deployed a recurrent neural network to design an iterative process for inference, by attending to one object at a time, and for each image, learning an appropriate number of inference steps. The authors showed that, in an un- supervised way, the proposed approach can learn generative models to identify multiple objects for both 2D and 3D problems.
Zhang and Zhu (2018) surveyed visual interpretability for deep learning.
5.4.2 RECOGNITION
RL can improve efï¬ciency for image classiï¬cation by focusing only on salient parts. For visual object localization and detection, RL can improve efï¬ciency over approaches with exhaustive spatial hypothesis search and sliding windows, and strikes a balance between sampling more regions for better accuracy and stopping the search when sufï¬cient conï¬dence is obtained about the targetâs location. | 1701.07274#161 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 162 | Mnih et al. (2014) introduced the recurrent attention model (RAM) to focus on selected sequence of regions or locations from an image or video for image classiï¬cation and object detection. The authors used REINFORCE to train the model, to overcome the issue that the model is non- differentiable, and experimented on an image classiï¬cation task and a dynamic visual control prob- lem.
Caicedo and Lazebnik (2015) proposed an active detection model for object localization with DQN, by deforming a bounding box with transformation actions to determine the most speciï¬c location for target objects. Jie et al. (2016) proposed a tree-structure RL approach to search for objects se- quentially, considering both the current observation and previous search paths, by maximizing the long-term reward associated with localization accuracy over all objects with DQN. Mathe et al. (2016) proposed to use policy search for visual object detection. Kong et al. (2017) deployed col- laborative multi-agent RL with inter-agent communication for joint object search. Welleck et al. (2017) proposed a hierarchical visual architecture with an attention mechanism for multi-label im- age classiï¬cation. Rao et al. (2017) proposed an attention-aware deep RL method for video face recognition. | 1701.07274#162 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 163 | Krull et al. (2017) for 6D object pose estimation
5.4.3 MOTION ANALYSIS
In tracking, an agent needs to follow a moving object. SupanËciËc and Ramanan (2017) proposed online decision-making process for tracking, formulated it as a partially observable decision-making process (POMDP), and learned policies with deep RL algorithms, to decide where to look for the object, when to reinitialize, and when to update the appearance model for the object, where image frames may be ambiguous and computational budget may be constrained. Yun et al. (2017) also studied visual tracking with deep RL.
Rhinehart and Kitani (2017) proposed Discovering Agent Rewards for K-futures Online (DARKO) to model and forecast ï¬rst-person camera wearerâs long-term goals, together with states, transitions, and rewards from streaming data, with online inverse reinforcement learning.
5.4.4 SCENE UNDERSTANDING
Wu et al. (2017b) studied the problem of scene understanding, and attempted to obtain a compact, expressive, and interpretable representation to encode scene information like objects, their cate- gories, poses, positions, etc, in a semi-supervised way. In contrast to encoder-decoder based neural
39 | 1701.07274#163 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 164 | 39
architectures as in previous works, Wu et al. (2017b) proposed to replace the decoder with a deter- ministic rendering function, to map a structured and disentangled scene description, scene XML, to an image; consequently, the encoder transforms an image to the scene XML by inverting the ren- dering operation, a.k.a., de-rendering. The authors deployed a variant of REINFORCE algorithm to overcome the non-differentiability issue of graphics rendering engines.
Wu et al. (2017a) proposed a paradigm with three major components, a convolutional perception module, a physics engine, and a graphics engine, to understand physical scenes without human an- notations. The perception module recovers a physical world representation by inverting the graphics engine, inferring the physical object state for each segment proposal in input and combining them. The generative physics and graphics engines then run forward with the world representation to re- construct the visual data. The authors showed results on both neural, differentiable and more mature but non-differentiable physics engines.
There are recent works about physics learning, e.g., Agrawal et al. (2016); Battaglia et al. (2016); Denil et al. (2017); Watters et al. (2017); Wu et al. (2015).
5.4.5 INTEGRATION WITH NLP | 1701.07274#164 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 165 | 5.4.5 INTEGRATION WITH NLP
Some are integrating computer vision with natural language processing. Xu et al. (2015) integrated attention to image captioning, trained the hard version attention with REINFORCE, and showed the effectiveness of attention on Flickr8k, Flickr30k, and MS COCO datasets. Rennie et al. (2017) introduced self-critical sequence training, using the output of test-time inference algorithm as the baseline in REINFORCE to normalize the rewards it experiences, for image captioning. See also Liu et al. (2016), Lu et al. (2016), and Ren et al. (2017) for image captioning. Strub et al. (2017) proposed end-to-end optimization with deep RL for goal-driven and visually grounded dialogue systems for GuessWhat?! game. Das et al. (2017) proposed to learn cooperative Visual Dialog agents with deep RL. See also Kottur et al. (2017). See Pasunuru and Bansal (2017) for video captioning. See Liang et al. (2017d) for visual relationship and attribute detection.
5.4.6 VISUAL CONTROL | 1701.07274#165 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 166 | 5.4.6 VISUAL CONTROL
Visual control is about deriving a policy from visual inputs, e.g., in games (Mnih et al., 2015; Silver et al., 2016a; 2017; Oh et al., 2015; Wu and Tian, 2017; Dosovitskiy and Koltun, 2017; Lample and Chaplot, 2017; Jaderberg et al., 2017), robotics (Finn and Levine, 2016; Gupta et al., 2017b; Lee et al., 2017; Levine et al., 2016a; Mirowski et al., 2017; Zhu et al., 2017b), and self-driving vehicles (Bojarski et al., 2016; Bojarski et al., 2017; Zhou and Tuzel, 2017). 3
5.5 BUSINESS MANAGEMENT
Reinforcement learning has many applications in business management, like ads, recommendation, customer management, and marketing.
Li et al. (2010) formulated personalized news articles recommendation as a contextual bandit prob- lem, to learn an algorithm to select articles sequentially for users based on contextual information of the user and articles, such as historical activities of the user and descriptive information and cate- gories of content, and to take user-click feedback to adapt article selection policy to maximize total user clicks in the long run. | 1701.07274#166 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 167 | Theocharous et al. (2015) formulated a personalized ads recommendation systems as a RL problem to maximize life-time value (LTV) with theoretical guarantees. This is in contrast to a myopic solution with supervised learning or contextual bandit formulation, usually with the performance metric of click through rate (CTR). As the models are hard to learn, the authors deployed a model- free approach to compute a lower-bound on the expected return of a policy to address the off-policy evaluation problem, i.e., how to evaluate a RL policy without deployment.
3Although we include visual control here, it is not very clear if we should categorize the following type of problems, e.g., DQN (Mnih et al., 2015) and AlphaGo (Silver et al., 2016a; 2017), into computer vision: pixels (DQN) or problem setting (Go board status) as the input, some deep neural networks as the architecture, and an end-to-end gradient descent/ascent algorithm as the optimization method to ï¬nd a policy, without any further knowledge of computer vision. Or we may see this as part of the synergy of computer vision and reinforcement learning.
40 | 1701.07274#167 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 168 | 40
Li et al. (2015) also attempted to maximize lifetime value of customers. Silver et al. (2013) pro- posed concurrent reinforcement learning for the customer interaction problem. See Sutton and Barto (2018) for a detailed and intuitive description of some topics discussed here under the section title of Personalized Web Services.
# 5.6 FINANCE
RL is a natural solution to some ï¬nance and economics problems (Hull, 2014; Luenberger, 1997), like option pricing (Longstaff and Schwartz, 2001; Tsitsiklis and Van Roy, 2001; Li et al., 2009), and multi-period portfolio optimization (Brandt et al., 2005; Neuneier, 1997), where value function based RL methods were used. Moody and Saffell (2001) proposed to utilize policy search to learn to trade; Deng et al. (2016) extended it with deep neural networks. Deep (reinforcement) learning would provide better solutions in some issues in risk management (Hull, 2014; Yu et al., 2009). The market efï¬ciency hypothesis is fundamental in ï¬nance. However, there are well-known behavioral biases in human decision-making under uncertainty, in particular, prospect theory (Prashanth et al., 2016). A reconciliation is the adaptive markets hypothesis (Lo, 2004), which may be approached by reinforcement learning. | 1701.07274#168 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 169 | It is nontrivial for ï¬nance and economics academia to accept blackbox methods like neural networks; Heaton et al. (2016) may be regarded as an exception. However, there is a lecture in AFA 2017 annual meeting: Machine Learning and Prediction in Economics and Finance. We may also be aware that ï¬nancial ï¬rms would probably hold state-of-the-art research/application results.
FinTech has been attracting attention, especially after the notion of big data. FinTech employs machine learning techniques to deal with issues like fraud detection (Phua et al., 2010), consumer credit risk (Khandani et al., 2010), etc.
# 5.7 HEALTHCARE | 1701.07274#169 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 170 | # 5.7 HEALTHCARE
There are many opportunities and challenges in healthcare for machine learning (Miotto et al., 2017; Saria, 2014). Personalized medicine is getting popular in healthcare. It systematically optimizes the patientâs health care, in particular, for chronic conditions and cancers using individual patient infor- mation, potentially from electronic health/medical record (EHR/EMR). Dynamic treatment regimes (DTRs) or adaptive treatment strategies are sequential decision making problems. Some issues in DTRs are not in standard RL. Shortreed et al. (2011) tackled the missing data problem, and designed methods to quantify the evidence of the learned optimal policy. Goldberg and Kosorok (2012) pro- posed methods for censored data (patients may drop out during the trial) and ï¬exible number of stages. See Chakraborty and Murphy (2014) for a recent survey, and Kosorok and Moodie (2015) for an edited book about recent progress in DTRs. Currently Q-learning is the RL method in DTRs. Ling et al. (2017) applied deep RL to the problem of inferring patient phenotypes. | 1701.07274#170 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 171 | the intersection of machine learning and healthcare are: NIPS Some recent workshops at 2016 Workshop on Machine Learning for Health (http://www.nipsml4hc.ws) and NIPS 2015 Workshop on Machine Learning in Healthcare (https://sites.google.com/site/nipsmlhc15/). See ICML 2017 Tutorial on Deep Learning for Health Care Applications: Challenges and Solutions (https://sites.google.com/view/icml2017-deep-health-tutorial/home).
5.8 EDUCATION
Mandel et al. (2014)
Liu et al. (2014)
5.9 INDUSTRY 4.0
The ear of Industry 4.0 is approaching, e.g., see OâDonovan et al. (2015), and Preuveneers and Ilie-Zudor (2017). Reinforcement learning in particular, artiï¬cial intelligence in general, will be critical enabling techniques for many aspects of Industry 4.0, e.g., predictive maintenance, real- time diagnostics, and management of manufacturing activities and processes. Robots will prevail in Industry 4.0, and we discuss robotics in Section 5.2.
41 | 1701.07274#171 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 172 | 41
Liu and Tomizuka (2016; 2017) studied how to make robots and people to collaborate to achieve both ï¬exibility and productivity in production lines. See a blog titled Towards Intelligent Industrial Co-robots, at http://bair.berkeley.edu/blog/2017/12/12/corobots/
Hein et al. (2017) designed a benchmark for the RL community to attempt to bridge the gap between academic research and real industrial problems. Its open source based on OpenAI Gym is available at https://github.com/siemens/industrialbenchmark.
Surana et al. (2016) proposed to apply guided policy search (Levine et al., 2016a) as discussed in Section 5.2.1 to optimize trajectory policy of cold spray nozzle dynamics, to handle complex trajec- tories traversing by robotic agents. The authors generated cold spray surface simulation proï¬les to train the model.
5.10 SMART GRID | 1701.07274#172 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 173 | 5.10 SMART GRID
A smart grid is a power grid utilizing modern information technologies to create an intelligent elec- tricity delivery network for electricity generation, transmission, distribution, consumption, and con- trol (Fang et al., 2012). An important aspect is adaptive control (Anderson et al., 2011). Glavic et al. (2017) reviewed application of RL for electric power system decision and control. Here we brieï¬y discuss demand response (Wen et al., 2015b; Ruelens et al., 2016). | 1701.07274#173 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 174 | Demand response systems motivate users to dynamically adapt electrical demands in response to changes in grid signals, like electricity price, temperature, and weather, etc. With suitable electricity prices, load of peak consumption may be rescheduled/lessened, to improve efï¬ciency, reduce costs, and reduce risks. Wen et al. (2015b) proposed to design a fully automated energy management system with model-free reinforcement learning, so that it doesnât need to specify a disutility function to model usersâ dissatisfaction with job rescheduling. The authors decomposed the RL formulation over devices, so that the computational complexity grows linearly with the number of devices, and conducted simulations using Q-learning. Ruelens et al. (2016) tackled the demand response problem with batch RL. Wen et al. (2015b) took the exogenous prices as states, and Ruelens et al. (2016) utilized the average as feature extractor to construct states.
INTELLIGENT TRANSPORTATION SYSTEMS
Intelligent transportation systems (Bazzan and Kl¨ugl, 2014) apply advanced information technolo- gies for tackling issues in transport networks, like congestion, safety, efï¬ciency, etc., to make trans- port networks, vehicles and users smart. | 1701.07274#174 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 175 | An important issue in intelligent transportation systems is adaptive trafï¬c signal control. El-Tantawy et al. (2013) proposed to model the adaptive trafï¬c signal control problem as a multiple player stochastic game, and solve it with the approach of multi-agent RL (Shoham et al., 2007; Busoniu et al., 2008). Multi-agent RL integrates single agent RL with game theory, facing challenges of stability, nonstationarity, and curse of dimensionality. El-Tantawy et al. (2013) approached the issue of coordination by considering agents at neighbouring intersections. The authors validated their proposed approach with simulations, and real trafï¬c data from the City of Toronto. El-Tantawy et al. (2013) didnât explore function approximation. See also van der Pol and Oliehoek (2017) for a recent work, and Mannion et al. (2016) for an experimental review, about applying RL to adaptive trafï¬c signal control.
Self-driving vehicle is also a topic of intelligent transportation systems. See Bojarski et al. (2016), Bojarski et al. (2017), Zhou and Tuzel (2017). | 1701.07274#175 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 176 | See NIPS 2017, 2016 Workshop on Machine Learning for Intelligent Transportation Systems. Check for a special issue of IEEE Transactions on Neural Networks and Learning Systems on Deep Re- inforcement Learning and Adaptive Dynamic Programming, tentative publication date December 2017.
5.12 COMPUTER SYSTEMS
Computer systems are indispensable in our daily life and work, e.g., mobile phones, computers, and cloud computing. Control and optimization problems abound in computer systems, e,g., Mestres
42
et al. (2016) proposed knowledge-deï¬ned networks, Gavrilovska et al. (2013) reviewed learning and reasoning techniques in cognitive radio networks, and Haykin (2005) discussed issues in cognitive radio, like channel state prediction and resource allocation. We also note that Internet of Things (IoT)(Xu et al., 2014) and wireless sensor networks (Alsheikh et al., 2014) play an important role in Industry 4.0 as discussed in Section 5.9, in Smart Grid as discussed in Section 5.10, and in Intelligent Transportation Systems as discussed in Section 5.11.
# 5.12.1 RESOURCE ALLOCATION | 1701.07274#176 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 177 | # 5.12.1 RESOURCE ALLOCATION
Mao et al. (2016) studied resource management in systems and networking with deep RL. The au- thors proposed to tackle multi-resource cluster scheduling with policy gradient, in an online manner with dynamic job arrivals, optimizing various objectives like average job slowdown or completion time. The authors validated their proposed approach with simulation.
Liu et al. (2017) proposed a hierarchical framework to tackle resource allocation and power man- agement in cloud computing with deep RL. The authors decomposed the problem as a global tier for virtual machines resource allocation and a local tier for servers power management. The au- thors validated their proposed approach with actual Google cluster traces. Such hierarchical frame- work/decomposition approach was to reduce state/action space, and to enable distributed operation of power management.
Google deployed machine learning for data centre power management, reducing energy consump- tion by 40%, https://deepmind.com/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-40/. Optimizing memory control is discussed in Sutton and Barto (2018).
5.12.2 PERFORMANCE OPTIMIZATION | 1701.07274#177 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 178 | 5.12.2 PERFORMANCE OPTIMIZATION
Mirhoseini et al. (2017) proposed to optimize device placement for Tensorï¬ow computational graphs with RL. The authors deployed a seuqence-to-sequence model to predict how to place subsets of operations in a Tensorï¬ow graph on available devices, using the execution time of the predicted placement as reward signal for REINFORCE algorithm. The proposed method found placements of Tensorï¬ow operations on devices for Inception-V3, recurrent neural language model and neural machine translation, yielding shorter execution time than those placements designed by human ex- perts. Computation burden is one concern for a RL approach to search directly in the solution space of a combinatorial problem. We discuss combinatorial optimization in Section ??.
5.12.3 SECURITY & PRIVACY
adversarial attacks, e.g., Huang et al. (2017),
http://rll.berkeley.edu/adversarial/
Papernot et al. (2016)
Abadi et al. (2016)
Balle et al. (2016)
Delle Fave et al. (2014)
Carlini and Wagner (2017b) | 1701.07274#178 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 179 | Abadi et al. (2016)
Balle et al. (2016)
Delle Fave et al. (2014)
Carlini and Wagner (2017b)
defenders Ian J. Goodfellow (2015); Carlini and Wagner (2017a); Madry et al. (2017); Tram`er et al. (2017)
Anderson et al. (2017) https://github.com/endgameinc/gym-malware
Evtimov et al. (2017) See a blog titled Physical Adversarial Examples Against Deep Neural Networks at http://bair.berkeley.edu/blog/2017/12/30/yolo-attack/, which contains a brief survey of attack and defence algorithms. Check ACM Conference on Computer and Communications Security (CCS 2016) tutorial on Adversarial Data Mining: Big Data Meets Cyber Security, https://www.sigsac.org/ccs/CCS2016/tutorials/index.html
43
# 6 MORE TOPICS | 1701.07274#179 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 180 | 43
# 6 MORE TOPICS
We list more interesting and/or important topics we have not discussed in this overview as below, hoping it would provide pointers for those who may be interested in studying them further. Some topics/papers may not contain RL yet. However, we believe these are interesting and/or important directions for RL in the sense of either theory or application. It would be deï¬nitely more desirable if we could ï¬nish reviewing these, however, we leave it as future work.
⢠understanding deep learning, Daniely et al. (2016); Li et al. (2016b); Karpathy et al. (2016); Kawaguchi et al. (2017); Koh and Liang (2017); Neyshabur et al. (2017); Shalev-Shwartz et al. (2017); Shwartz-Ziv and Tishby (2017); Zhang et al. (2017)
⢠interpretability, e.g., Al-Shedivat et al. (2017b); Doshi-Velez and Kim (2017); Harrison et al. (2017); Lei et al. (2016); Lipton (2016); Miller (2017); Ribeiro et al. (2016); Huk Park et al. (2016) | 1701.07274#180 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 181 | â NIPS 2017 Interpretable Machine Learning Symposium â ICML 2017 Tutorial on Interpretable Machine Learning â NIPS 2016 Workshop on Interpretable ML for Complex Systems â ICML Workshop on Human Interpretability in Machine Learning 2017, 2016
usable machine learning, Bailis et al. (2017)
⦠Cloud AutoML: Making AI accessible to every business, https://www.blog.google/topics/google-cloud/cloud-automl-making-ai-accessible-every- business/
expressivity, Raghu et al. (2016) ⢠testing, Pei et al. (2017) ⢠deep learning efï¬ciency, e.g., Han et al. (2016), Spring and Shrivastava (2017), Sze et al.
(2017) | 1701.07274#181 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 182 | (2017)
deep learning compression ⢠optimization, e.g., Wilson et al. (2017), Czarnecki et al. (2017) ⢠normalization, Klambauer et al. (2017), van Hasselt et al. (2016b) ⢠curriculum learning, Graves et al. (2017), Held et al. (2017), Matiisen et al. (2017) ⢠professor forcing, Lamb et al. (2016) ⢠new Q-value operators, Kavosh and Littman (2017), Haarnoja et al. (2017) ⢠large action space, e.g., Dulac-Arnold et al. (2015); He et al. (2016c) ⢠Predictive State Representation, Downey et al. (2017), Venkatraman et al. (2017) ⢠safe RL, e.g., Berkenkamp et al. (2017) ⢠agent modelling, e.g., Albrechta and Stone (2018) ⢠semi-supervised learning, e.g., Audiffren et al. (2015); Cheng et al. (2016); Dai et al. (2017); Finn et al. (2017); Kingma et al. (2014); Papernot et al. (2017); Yang et al. (2017); Zhu and Goldberg (2009) | 1701.07274#182 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 183 | neural episodic control, Pritzel et al. (2017) ⢠continual learning, Chen and Liu (2016); Kirkpatrick et al. (2017); Lopez-Paz and Ranzato
(2017)
â Satinder Singh, Steps towards continual learning, tutorial at Deep Learning and Rein- forcement Learning Summer School 2017
⢠symbolic learning, Evans and Grefenstette (2017); Liang et al. (2017a); Parisotto et al. (2017)
pathNet, Fernando et al. (2017) ⢠evolution strategies, Petroski Such et al. (2017), Salimans et al. (2017) , Lehman et al.
(2017)
44
capsules, Sabour et al. (2017) ⢠DeepForest, Zhou and Feng (2017); Feng and Zhou (2017) ⢠deep probabilistic programming, Tran et al. (2017) ⢠active learning, e.g., Fang et al. (2017) ⢠deep learning games, Schuurmans and Zinkevich (2016) ⢠program learning, e.g., Balog et al. (2017); Cai et al. (2017); Denil et al. (2017); Parisotto
et al. (2017); Reed and de Freitas (2016) | 1701.07274#183 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 184 | et al. (2017); Reed and de Freitas (2016)
relational reasoning, e.g., Santoro et al. (2017), Watters et al. (2017) ⢠proving, e.g., Loos et al. (2017); Rockt¨aschel and Riedel (2017) ⢠education, e.g., ⢠music generation, e.g., Jaques et al. (2017) ⢠retrosynthesis, e.g., Segler et al. (2017) ⢠quantum RL, e.g., Crawford et al. (2016)
# â NIPS 2015 Workshop on Quantum Machine Learning
# 7 RESOURCES
We list a collection of deep RL resources including books, surveys, reports, online courses, tutorials, conferences, journals and workshops, blogs, testbed, and open source algorithm implementations. This by no means is complete.
It is essential to have a good understanding of reinforcement learning, before having a good under- standing of deep reinforcement learning. We recommend to start with the textbook by Sutton and Barto (Sutton and Barto, 2018), the RL courses by Rich Sutton and by David Silver as the ï¬rst two items in the Courses subsection below. | 1701.07274#184 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 185 | In the current information/social media age, we are overwhelmed by information, e.g., from Twitter, arXiv, Google+, etc. The skill to efï¬ciently select the best information becomes essential. The Wild Week in AI (http://www.wildml.com) is an excellent series of weekly summary blogs. In an ear of AI, we expect to see an AI agent to do such tasks like intelligently searching and summarizing relevant news, blogs, research papers, etc.
7.1 BOOKS
⢠the deï¬nitive and intuitive reinforcement learning book by Richard S. Sutton and Andrew G. Barto (Sutton and Barto, 2018)
⢠deep learning books (Deng and Dong, 2014; Goodfellow et al., 2016)
7.2 MORE BOOKS | 1701.07274#185 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 186 | ⢠deep learning books (Deng and Dong, 2014; Goodfellow et al., 2016)
7.2 MORE BOOKS
theoretical RL books (Bertsekas, 2012; Bertsekas and Tsitsiklis, 1996; Szepesv´ari, 2010) ⢠an operations research oriented RL book (Powell, 2011) ⢠an edited RL book (Wiering and van Otterlo, 2012) ⢠Markov decision processes (Puterman, 2005) ⢠machine learning (Bishop, 2011; Hastie et al., 2009; Haykin, 2008; James et al., 2013; Kuhn and Johnson, 2013; Murphy, 2012; Provost and Fawcett, 2013; Simeone, 2017; Zhou, 2016)
artiï¬cial intelligence (Russell and Norvig, 2009) ⢠natural language processing (NLP) (Deng and Liu, 2017; Goldberg, 2017; Jurafsky and
Martin, 2017)
semi-supervised learning (Zhu and Goldberg, 2009) ⢠game theory (Leyton-Brown and Shoham, 2008)
45
7.3 SURVEYS AND REPORTS
⢠reinforcement learning (Littman, 2015; Kaelbling et al., 1996; Geramifard et al., 2013; Grondman et al., 2012; Roijers et al., 2013); deep reinforcement learning (Arulkumaran et al., 2017) 4 | 1701.07274#186 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 187 | deep learning (LeCun et al., 2015; Schmidhuber, 2015; Bengio, 2009; Wang and Raj, 2017) ⢠efï¬cient processing of deep neural networks (Sze et al., 2017) ⢠machine learning (Jordan and Mitchell, 2015) ⢠practical machine learning advices (Domingos, 2012; Smith, 2017; Zinkevich, 2017) ⢠natural language processing (NLP) (Hirschberg and Manning, 2015; Cho, 2015; Young
et al., 2017)
⢠spoken dialogue systems (Deng and Li, 2013; Hinton et al., 2012; He and Deng, 2013; Young et al., 2013)
robotics (Kober et al., 2013) ⢠transfer learning (Taylor and Stone, 2009; Pan and Yang, 2010; Weiss et al., 2016) ⢠Bayesian RL (Ghavamzadeh et al., 2015) ⢠AI safety (Amodei et al., 2016; Garc`ıa and Fern`andez, 2015) ⢠Monte Carlo tree search (MCTS) (Browne et al., 2012; Gelly et al., 2012)
7.4 COURSES
⢠Richard Sutton, Reinforcement Learning, 2016, slides, assignments, reading materials, etc. http://www.incompleteideas.net/sutton/609%20dropbox/ | 1701.07274#187 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 188 | ⢠Richard Sutton, Reinforcement Learning, 2016, slides, assignments, reading materials, etc. http://www.incompleteideas.net/sutton/609%20dropbox/
⢠David Silver, Reinforcement Learning, 2015, slides (goo.gl/UqaxlO), video-lectures (goo.gl/7BVRkT)
⢠Sergey Levine, John Schulman and Chelsea Finn, CS 294: Deep Reinforcement Learning, Spring 2017, http://rll.berkeley.edu/deeprlcourse/
⢠Katerina Fragkiadaki, Ruslan Satakhutdinov, Deep Reinforcement Learning and Control, Spring 2017, https://katefvision.github.io
Emma Brunskill, CS234: Reinforcement Learning, http://web.stanford.edu/class/cs234/ ⢠Charles Isbell, Michael Littman and Pushkar Kolhe, Udacity: Machine Learning: Reinforcement Learning, goo.gl/eyvLfg
⢠David Donoho, Hatef Monajemi, and Vardan Papyan, Theories of Deep Learning (Stanford STATS 385), https://stats385.github.io
Nando de Freitas, Deep Learning Lectures, https://www.youtube.com/user/ProfNandoDF ⢠Fei-Fei Li, Andrej Karpathy and Justin Johnson, CS231n: Convolutional Neural Networks | 1701.07274#188 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 189 | for Visual Recognition, http://cs231n.stanford.edu
⢠Richard Socher, CS224d: Deep Learning for Natural Language Processing, http://cs224d.stanford.edu
⢠Brendan Shillingford, Yannis Assael, Chris Dyer, Oxford Deep NLP 2017 course, https://github.com/oxford-cs-deepnlp-2017
⢠Pieter Abbeel, Advanced Robotics, Fall 2015, https://people.eecs.berkeley.edu/ pabbeel/cs287- fa15/
⢠Emo Todorov, Intelligent control through learning and optimization, http://homes.cs.washington.edu/â¼todorov/courses/amath579/index.html
Abdeslam Boularias, Robot Learning Seminar, http://www.abdeslam.net/robotlearningseminar ⢠MIT 6.S094: Deep Learning for Self-Driving Cars, http://selfdrivingcars.mit.edu ⢠Jeremy Howard, Practical Deep Learning For Coders, http://course.fast.ai ⢠Andrew Ng, Deep Learning Specialization
https://www.coursera.org/specializations/deep-learning
4Our overview is much more comprehensive, and was online much earlier, than this brief survey.
46
7.5 TUTORIALS | 1701.07274#189 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 190 | 4Our overview is much more comprehensive, and was online much earlier, than this brief survey.
46
7.5 TUTORIALS
⢠Rich Sutton, Introduction to Reinforcement Learning with Function Approximation, https://www.microsoft.com/en-us/research/video/tutorial-introduction-to-reinforcement- learning-with-function-approximation/
Deep Reinforcement Learning â David Silver, ICML 2016 â David Silver, 2nd Multidisciplinary Conference on Reinforcement Learning and Decision Making (RLDM), Edmonton, Alberta, Canada, 2015; http://videolectures.net/rldm2015 silver reinforcement learning/
â John Schulman, Deep Learning School, 2016 â Pieter Abbeel, Deep Learning Summer School, 2016;
# http://videolectures.net/deeplearning2016 abbeel deep reinforcement/
â Pieter Abbeel and John Schulman, Deep Reinforcement Learning Through Policy Op- timization, NIPS 2016
â Sergey Levine and Chelsea Finn, Deep Reinforcement Learning, Decision Making, and Control, ICML 2017
⢠John Schulman, The Nuts and Bolts of Deep Reinforcement Learning Research, Deep Re- inforcement Learning Workshop, NIPS 2016
⢠Joelle Pineau, Introduction to Reinforcement Learning, Deep Learning Summer School, 2016; http://videolectures.net/deeplearning2016 pineau reinforcement learning/ | 1701.07274#190 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 191 | ⢠Joelle Pineau, Introduction to Reinforcement Learning, Deep Learning Summer School, 2016; http://videolectures.net/deeplearning2016 pineau reinforcement learning/
Andrew Ng, Nuts and Bolts of Building Applications using Deep Learning, NIPS 2016 ⢠Deep Learning Summer School, 2016, 2015 ⢠Deep Learning and Reinforcement Learning Summer Schools, 2017 ⢠Simons Institute Interactive Learning Workshop, 2017 ⢠Simons Institute Representation Learning Workshop, 2017 ⢠Simons Institute Computational Challenges in Machine Learning Workshop, 2017
7.6 CONFERENCES, JOURNALS AND WORKSHOPS | 1701.07274#191 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 192 | 7.6 CONFERENCES, JOURNALS AND WORKSHOPS
NIPS: Neural Information Processing Systems ⢠ICML: International Conference on Machine Learning ⢠ICLR: International Conference on Learning Representation ⢠RLDM: Multidisciplinary Conference on Reinforcement Learning and Decision Making ⢠EWRL: European Workshop on Reinforcement Learning ⢠AAAI, IJCAI, ACL, EMNLP, SIGDIAL, ICRA, IROS, KDD, SIGIR, CVPR, etc. ⢠Nature Machine Intelligence, Science Robotics, JMLR, MLJ, AIJ, JAIR, PAMI, etc ⢠Nature May 2015, Science July 2015, survey papers on machine learning/AI ⢠Science, July 7, 2017 issue, The Cyberscientist, a special issue about AI ⢠Deep Reinforcement Learning Workshop, NIPS 2016, 2015; IJCAI 2016 ⢠Deep Learning Workshop, ICML 2016 ⢠http://distill.pub
7.7 BLOGS
⢠Deepmind Blog, https://deepmind.com/blog/ | 1701.07274#192 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 193 | ⢠Google Research Blog, https://research.googleblog.com, goo.gl/ok88b7
⦠The Google Brain Team â Looking Back on 2017, goo.gl/1G7jnb, goo.gl/uCWDLr
47
The Google Brain Team â Looking Back on 2016, ⢠Berkeley AI Research Blog, http://bair.berkeley.edu/blog/ ⢠OpenAI Blog, https://blog.openai.com ⢠Marc Bellemare, Classic and modern reinforcement learning,
http://www.marcgbellemare.info/blog/
⢠Denny Britz, The Wild Week in AI, a weekly AI & deep learning newsletter, www.wildml.com, esp. goo.gl/MyrwDC
Andrej Karpathy, karpathy.github.io, esp. goo.gl/1hkKrb ⢠Junling Hu, Reinforcement learning explained - learning to act based on long-term payoffs
https://www.oreilly.com/ideas/reinforcement-learning-explained
⢠Li Deng, How deep reinforcement learning can help chatbots | 1701.07274#193 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 194 | https://www.oreilly.com/ideas/reinforcement-learning-explained
⢠Li Deng, How deep reinforcement learning can help chatbots
# https://venturebeat.com/2016/08/01/how-deep-reinforcement-learning-can-help-chatbots/ ⢠Reinforcement Learning, https://www.technologyreview.com/s/603501/10-breakthroughtechnologies-2017-reinforcement-learning/
Deep Learning, https://www.technologyreview.com/s/513696/deep-learning/
7.8 TESTBEDS
⢠The Arcade Learning Environment (ALE) (Bellemare et al., 2013; Machado et al., 2017) is a framework composed of Atari 2600 games to develop and evaluate AI agents.
⢠Ray RLlib: A Composable and Scalable Reinforcement Learning Library (Liang et al., 2017c), http://ray.readthedocs.io/en/latest/rllib.html
⢠OpenAI Gym (https://gym.openai.com) is a toolkit for the development of RL algorithms, consisting of environments, e.g., Atari games and simulated robots, and a site for the com- parison and reproduction of results. | 1701.07274#194 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 195 | ⢠OpenAI Universe (https://universe.openai.com) is used to turn any program into a Gym environment. Universe has already integrated many environments, including Atari games, ï¬ash games, browser tasks like Mini World of Bits and real-world browser tasks. Recently, GTA V was added to Universe for self-driving vehicle simulation.
DeepMind Control Suite, Tassa et al. (2018) ⢠DeepMind released a ï¬rst-person 3D game platform DeepMind Lab (Beattie et al., 2016). Deepmind and Blizzard will collaborate to release the Starcraft II AI research environment (goo.gl/Ptiwfg).
⢠Psychlab: A Psychology Laboratory for Deep Reinforcement Learning Agents (Leibo et al., 2018)
⢠FAIR TorchCraft (Synnaeve et al., 2016) is a library for Real-Time Strategy (RTS) games such as StarCraft: Brood War.
Deepmind PySC2 - StarCraft II Learning Environment, https://github.com/deepmind/pysc2 ⢠David Churchill, CommandCenter: StarCraft 2 AI Bot,
https://github.com/davechurchill/commandcenter
⢠ParlAI is a framework for dialogue research, implemented in Python, open-sourced by Facebook. https://github.com/facebookresearch/ParlAI | 1701.07274#195 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 196 | ⢠ParlAI is a framework for dialogue research, implemented in Python, open-sourced by Facebook. https://github.com/facebookresearch/ParlAI
ELF, an extensive, lightweight and ï¬exible platform for RL research (Tian et al., 2017) ⢠Project Malmo (https://github.com/Microsoft/malmo), from Microsoft, is an AI research
and experimentation platform built on top of Minecraft.
Twitter open-sourced torch-twrl, a framework for RL development. ⢠ViZDoom is a Doom-based AI research platform for visual RL (Kempka et al., 2016). ⢠Baidu Apollo Project, self-driving open-source, http://apollo.auto ⢠TORCS is a car racing simulator (Bernhard Wymann et al., 2014). ⢠MuJoCo, Multi-Joint dynamics with Contact, is a physics engine, http://www.mujoco.org.
48
Nogueira and Cho (2016) presented WebNav Challenge for Wikipedia links navigation. ⢠RLGlue (Tanner and White, 2009) is a language-independent software for RL experiments.
It may need extensions to accommodate progress in deep learning.
⢠RLPy (Geramifard et al., 2015) is a value-function-based reinforcement learning frame- work for education and research.
7.9 ALGORITHM IMPLEMENTATIONS | 1701.07274#196 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 197 | 7.9 ALGORITHM IMPLEMENTATIONS
We collect implementations of algorithms, either classical ones as in a textbook like Sutton and Barto (2018) or in recent papers.
⢠Shangtong Zhang, Python code to accompany Sutton & Bartoâs RL book and David Silverâs RL course, https://github.com/ShangtongZhang/reinforcement-learning-an-introduction
⢠Learning Reinforcement Learning (with Code, Exercises and Solutions), http://www.wildml.com/2016/10/learning-reinforcement-learning/
⢠OpenAI Baselines: high-quality implementations of reinforcement learning algorithms, https://github.com/openai/baselines
⢠TensorFlow implementation of Deep Reinforcement Learning papers, https://github.com/carpedm20/deep-rl-tensorï¬ow
Deep reinforcement learning for Keras, https://github.com/matthiasplappert/keras-rl ⢠Code Implementations for NIPS 2016 papers, http://bit.ly/2hSaOyx ⢠Benchmark results of various policy optimization algorithms (Duan et al., 2016),
https://github.com/rllab/rllab | 1701.07274#197 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 198 | https://github.com/rllab/rllab
Tensor2Tensor (T2T) (Vaswani et al., 2017; Kaiser et al., 2017a;b) ⢠DQN (Mnih et al., 2015), https://sites.google.com/a/deepmind.com/dqn/ ⢠Tensorï¬ow implementation of DQN (Mnih et al., 2015),
https://github.com/devsisters/DQN-tensorï¬ow
Deep Q Learning with Keras and Gym, https://keon.io/deep-q-learning/ ⢠Deep Exploration via Bootstrapped DQN (Osband et al., 2016), a Torch implementation,
https://github.com/iassael/torch-bootstrapped-dqn
DarkForest, the Facebook Go engine (Github), https://github.com/facebookresearch/darkforestGo ⢠Using Keras and Deep Q-Network to Play FlappyBird,
https://yanpanlau.github.io/2016/07/10/FlappyBird-Keras.html
⢠Deep Deterministic Policy Gradients (Lillicrap et al., 2016) in TensorFlow, http://pemami4911.github.io/blog/2016/08/21/ddpg-rl.html | 1701.07274#198 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 199 | ⢠Deep Deterministic Policy Gradient (Lillicrap et al., 2016) to play TORCS, https://yanpanlau.github.io/2016/10/11/Torcs-Keras.html
⢠Reinforcement learning with unsupervised auxiliary tasks (Jaderberg et al., 2017), https://github.com/miyosuda/unreal
⢠Learning to communicate with deep multi-agent reinforcement learning, https://github.com/iassael/learning-to-communicate
Deep Reinforcement Learning: Playing a Racing Game - Byte Tank, http://bit.ly/2pVIP4i ⢠Differentiable Neural Computer (DNC) (Graves et al., 2016),
https://github.com/deepmind/dnc
⢠Playing FPS Games with Deep Reinforcement Learning (Lample and Chaplot, 2017), https://github.com/glample/Arnold
⢠Learning to Learn (Reed and de Freitas, 2016) in TensorFlow, https://github.com/deepmind/learning-to-learn
⢠Value Iteration Networks (Tamar et al., 2016) in Tensorï¬ow, https://github.com/TheAbhiKumar/tensorï¬ow-value-iteration-networks
49 | 1701.07274#199 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 200 | 49
⢠Tensorï¬ow implementation of the Predictron (Silver et al., 2016b), https://github.com/zhongwen/predictron
⢠Meta Reinforcement Learning (Wang et al., 2016) in Tensorï¬ow, https://github.com/awjuliani/Meta-RL
⢠Generative adversarial imitation learning (Ho and Ermon, 2016), containing an im- plementation of Trust Region Policy Optimization (TRPO) (Schulman et al., 2015), https://github.com/openai/imitation
⢠Starter code for evolution strategies (Salimans et al., 2017), https://github.com/openai/evolution-strategies-starter
Transfer learning (Long et al., 2015; 2016), https://github.com/thuml/transfer-caffe ⢠DeepForest (Zhou and Feng, 2017), http://lamda.nju.edu.cn/ï¬les/gcforest.zip
# 8 BRIEF SUMMARY
We list some RL issues and corresponding proposed approaches covered in this overview, as well as some classical work. One direction of future work is to further reï¬ne this section, especially for issues and solutions in applications.
issue: prediction, policy evaluation
proposed approaches: | 1701.07274#200 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 201 | issue: prediction, policy evaluation
proposed approaches:
â temporal difference (TD) learning (Sutton, 1988) ⢠issue: control, ï¬nding optimal policy (classical work)
proposed approaches:
# â Q-learning (Watkins and Dayan, 1992) â policy gradient (Williams, 1992)
© reduce variance of gradient estimate: baseline, advantage function (Williams, 1992; Sutton et al., 2000)
# â actor-critic (Barto et al., 1983) â SARSA (Sutton and Barto, 2018)
⢠issue: the deadly triad: instability and divergence when combining off-policy, function
approximation, and bootstrapping proposed approaches:
â DQN with experience replay (Lin, 1992) and target network (Mnih et al., 2015) © overestimate problem in Q-learning: double DQN (van Hasselt et al., 201 6a) © prioritized experience replay (Schaul et al., 2016) © better exploration strategy (Osband et al., 2016) © optimality tightening to accelerate DQN (He et al., 2017) © reduce variability and instability with averaged-DQN (Anschel et al., 2017) | 1701.07274#201 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 202 | â dueling architecture (Wang et al., 2016b) â asynchronous methods (Mnih et al., 2016) â trust region policy optimization (Schulman et al., 2015) â distributed proximal policy optimization (Heess et al., 2017) â combine policy gradient and Q-learning (OâDonoghue et al., 2017; Nachum et al.,
2017; Gu et al., 2017; Schulman et al., 2017)
â GTD (Sutton et al., 2009a;b; Mahmood et al., 2014) â Emphatic-TD (Sutton et al., 2016)
issue: train perception and control jointly end-to-end
proposed approaches:
â guided policy search (Levine et al., 2016a)
issue: data/sample efï¬ciency
50
proposed approaches:
â Q-learning, actor-critic â model-based policy search, e.g., PILCO Deisenroth and Rasmussen (2011) â actor-critic with experience replay (Wang et al., 2017b) â PGQ, policy gradient and Q-learning (OâDonoghue et al., 2017) â Q-Prop, policy gradient with off-policy critic (Gu et al., 2017) â return-based off-policy control, Retrace (Munos et al., 2016), Reactor (Gruslys et al.,
2017) | 1701.07274#202 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 203 | 2017)
â learning to learn, e.g., Duan et al. (2016); Wang et al. (2016); Lake et al. (2015)
issue: reward function not available
proposed approaches:
â imitation learning â inverse RL (Ng and Russell, 2000) â learn from demonstration (Hester et al., 2018) â imitation learning with GANs (Ho and Ermon, 2016; Stadie et al., 2017) â train dialogue policy jointly with reward model (Su et al., 2016b)
issue: exploration-exploitation tradeoff
proposed approaches:
â unify count-based exploration and intrinsic motivation (Bellemare et al., 2016) â under-appreciated reward exploration (Nachum et al., 2017) â deep exploration via bootstrapped DQN (Osband et al., 2016) â variational information maximizing exploration (Houthooft et al., 2016)
issue: model-based learning
proposed approaches:
â Dyna-Q (Sutton, 1990) â combine model-free and model-based RL (Chebotar et al., 2017)
issue: model-free planning
proposed approaches:
â value iteration networks (Tamar et al., 2016) â predictron (Silver et al., 2016b)
issue: focus on salient parts
proposed approaches: attention | 1701.07274#203 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 204 | proposed approaches:
â value iteration networks (Tamar et al., 2016) â predictron (Silver et al., 2016b)
issue: focus on salient parts
proposed approaches: attention
â object detection (Mnih et al., 2014) â neural machine translation (Bahdanau et al., 2015) â image captioning (Xu et al., 2015) â replace CNN and RNN with attention in sequence modelling (Vaswani et al., 2017)
issue: data storage over long time, separating from computation
proposed approaches: memory
# â differentiable neural computer (DNC) with external memory (Graves et al., 2016)
issue: beneï¬t from non-reward training signals in environments
proposed approaches: unsupervised Learning
â Horde (Sutton et al., 2011) â unsupervised reinforcement and auxiliary learning (Jaderberg et al., 2017) â learn to navigate with unsupervised auxiliary learning (Mirowski et al., 2017) â generative adversarial networks (GANs) (Goodfellow et al., 2014)
⢠issue: learn knowledge from different domains
51
proposed approaches: Weiss et al., 2016) transfer Learning (Taylor and Stone, 2009; Pan and Yang, 2010;
â learn invariant features to transfer skills (Gupta et al., 2017a)
issue: beneï¬t from both labelled and unlabelled data | 1701.07274#204 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 205 | â learn invariant features to transfer skills (Gupta et al., 2017a)
issue: beneï¬t from both labelled and unlabelled data
proposed approaches: semi-supervised learning (Zhu and Goldberg, 2009)
â learn with MDPs both with and without reward functions (Finn et al., 2017) â learn with expertâs trajectories and those may not from experts (Audiffren et al., 2015) ⢠issue: learn, plan, and represent knowledge with spatio-temporal abstraction at multiple
levels
proposed approaches: hierarchical RL (Barto and Mahadevan, 2003)
â options (Sutton et al., 1999), MAXQ (Dietterich, 2000) â strategic attentive writer to learn macro-actions (Vezhnevets et al., 2016) â integrate temporal abstraction with intrinsic motivation (Kulkarni et al., 2016) â stochastic neural networks for hierarchical RL (Florensa et al., 2017) â lifelong learning with hierarchical RL (Tessler et al., 2017)
issue: adapt rapidly to new tasks
proposed approaches: learning to learn
â learn to optimize (Li and Malik, 2017) â learn a ï¬exible RNN model to handle a family of RL tasks (Duan et al., 2016; Wang
et al., 2016) | 1701.07274#205 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 206 | et al., 2016)
â one/few/zero-shot learning (Duan et al., 2017; Johnson et al., 2016; Kaiser et al., 2017b; Koch et al., 2015; Lake et al., 2015; Ravi and Larochelle, 2017; Vinyals et al., 2016)
issue: gigantic search space
proposed approaches:
â integrate supervised learning, reinforcement learning, and Monte-Carlo tree search as in AlphaGo (Silver et al., 2016a)
issue: neural networks architecture design
proposed approaches:
â neural architecture search (Bello et al., 2017; Baker et al., 2017; Zoph and Le, 2017) â new architectures, e.g., Kaiser et al. (2017a), Silver et al. (2016b), Tamar et al. (2016),
Vaswani et al. (2017), Wang et al. (2016b)
# 9 DISCUSSIONS
It is both the best and the worst of times for the ï¬eld of deep RL, for the same reason: it has been growing so fast and so enormously. We have been witnessing breakthroughs, exciting new methods and applications, and we expect to see much more and much faster. As a consequence, this overview is incomplete, in the sense of both depth and width. However, we attempt to summarize important achievements and discuss potential directions and applications in this amazing ï¬eld. | 1701.07274#206 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 207 | In this overview, we summarize six core elements â value function, policy, reward, model and plan- ning, exploration, and knowledge; six important mechanisms â attention and memory, unsupervised learning, transfer learning, multi-agent RL, hierarchical RL, and learning to learn; and twelve ap- plications â games, robotics, natural language processing, computer vision, business management, ï¬nance, healthcare, education, Industry 4.0, smart grid, intelligent transportation systems, and com- puter systems. We also discuss background of machine learning, deep learning, and reinforcement learning, and list a collection of RL resources.
We have seen breakthroughs about deep RL, including deep Q-network (Mnih et al., 2015) and AlphaGo (Silver et al., 2016a). There have been many extensions to, improvements for and applica- tions of deep Q-network (Mnih et al., 2015).
52 | 1701.07274#207 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 208 | 52
Novel architectures and applications using deep RL were recognized in top tier conferences as best papers in 2016: dueling network architectures (Wang et al., 2016b) at ICML, spoken dialogue systems (Su et al., 2016b) at ACL (student), information extraction (Narasimhan et al., 2016) at EMNLP, and, value iteration networks (Tamar et al., 2016) at NIPS. Gelly and Silver (2007) was the recipient of Test of Time Award at ICML 2017. In 2017, the following were recognized as best papers, Kottur et al. (2017) at EMNLP (short), and, Bacon et al. (2017) at AAAI (student). Exciting achievements abound: differentiable neural computer (Graves et al., 2016), unsupervised reinforcement and auxiliary learning (Jaderberg et al., 2017), asynchronous methods (Mnih et al., 2016), dual learning for machine translation (He et al., 2016a), guided policy search (Levine et al., 2016a), generative adversarial imitation learning (Ho and Ermon, 2016), and neural architecture de- sign (Zoph and Le, 2017), etc. Creativity would push the frontiers of deep RL further with respect to core elements, mechanisms, and applications.
State of the Art Control of Atari Games Using Shallow RL was accepted at AAMAS. It was also nominated for the Best Paper Award | 1701.07274#208 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 209 | State of the Art Control of Atari Games Using Shallow RL was accepted at AAMAS. It was also nominated for the Best Paper Award
Value function is central to reinforcement learning, e.g., in deep Q-network and its many exten- tions. Policy optimization approaches have been gaining traction, in many, diverse applications, e.g., robotics, neural architecture design, spoken dialogue systems, machine translation, attention, and learning to learn, and this list is boundless. New learning mechanisms have emerged, e.g., using transfer/unsupervised/semi-supervised learning to improve the quality and speed of learn- ing, and more new mechanisms will be emerging. This is the renaissance of reinforcement learn- ing (Krakovsky, 2016). In fact, reinforcement learning and deep learning have been making steady progress even during the last AI winter.
A popular criticism about deep learning is that it is a blackbox, or even the âalchemyâ as a com- ment during NIPS 2017 Test of Time Award (Rahimi and Recht, 2007) speech, so it is not clear how it works. This should not be the reason not to accept deep learning; rather, having a better un- derstanding of how deep learning works is helpful for deep learning and general machine learning community. There are works in this direction as well as for interpretability of deep learning as we list in Section 6. | 1701.07274#209 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 210 | It is essential to consider issues of learning models, like stability, convergence, accuracy, data efï¬- ciency, scalability, speed, simplicity, interpretability, robustness, and safety, etc. It is important to investigate comments/criticisms, e.g., from conginitive science, like intuitive physics, intuitive psy- chology, causal model, compositionality, learning to learn, and act in real time (Lake et al., 2016), for stronger AI. It is interesting to check Deepmindâs commentary (Botvinick et al., 2017) about one additional ingredient, autonomy, so that agents can build and exploit their own internal models, with minimal human manual engineering, and investigate the connection between neuroscience and RL/AI (Hassabis et al., 2017). See also Peter Norvigâs perspective at http://bit.ly/2qpehcd. See Sto- ica et al. (2017) for systems challenges for AI. | 1701.07274#210 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 211 | Nature in May 2015 and Science in July 2015 featured survey papers on machine learning/AI. Sci- ence Robotics launched in 2016. Science has a special issue on July 7, 2017 about AI on The Cyberscientist. Nature Machine Intelligence will launch in January 2019. The coverage of AI by premier journals like Nature and Science and the launch of Science Robotics and Nature Machine Intelligence illustrate the apparent importance of AI. It is interesting to mention that NIPS 2017 main conference was sold out only two weeks after opening for registration.
It is worthwhile to envision deep RL considering perspectives of government, academia and industry on AI, e.g., Artiï¬cial Intelligence, Automation, and the economy, Executive Ofï¬ce of the President, USA; Artiï¬cial Intelligence and Life in 2030 - One Hundred Year Study on Artiï¬cial Intelligence: Report of the 2015-2016 Study Panel, Stanford University (Stone et al., 2016); and AI, Machine Learning and Data Fuel the Future of Productivity by The Goldman Sachs Group, Inc., etc. See also the recent AI Frontiers Conference, https://www.aifrontiers.com. | 1701.07274#211 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 212 | Deep learning was among MIT Technology Review 10 Breakthrough Technologies in 2013. We have been witnessing the dramatic development of deep learning in both academia and industry in the last few years. Reinforcement learning was among MIT Technology Review 10 Breakthrough Technologies in 2017. Deep learning has made many achievements, has âconqueredâ speech recog- nition, computer vision, and now NLP, is more mature and well-accepted, and has been validated by products and market. In contrast, RL has lots of (potential, promising) applications, yet few products
53
so far, may still need better algorithms, may still need products and market validation. However, it is probably the right time to nurture, educate and lead the market. We will see both deep learning and reinforcement learning prospering in the coming years and beyond. Prediction is very difï¬cult, especially about the future. However, we expect that 2018 for reinforcement learning would be 2010 for deep learning. | 1701.07274#212 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 213 | Deep learning, in this third wave of AI, will have deeper inï¬uences, as we have already seen from its many achievements. Reinforcement learning, as a more general learning and decision making paradigm, will deeply inï¬uence deep learning, machine learning, and artiï¬cial intelligence in gen- eral. Deepmind, conducting leading research in deep reinforcement learning, recently opened its ï¬rst ever international AI research ofï¬ce in Alberta, Canada, co-locating with the major research center for reinforcement learning led by Rich Sutton. It is interesting to mention that when Pro- fessor Rich Sutton started working in the University of Alberta in 2003, he named his lab RLAI: Reinforcement Learning and Artiï¬cial Intelligence.
# ACKOWLEDGEMENT | 1701.07274#213 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 214 | # ACKOWLEDGEMENT
We appreciate comments from Baochun Bai, Kan Deng, Hai Fang, Hua He, Junling Hu, Ruitong Huang, Aravind Lakshminarayanan, Jinke Li, Lihong Li, Bhairav Mehta, Dale Schuurmans, David Silver, Rich Sutton, Csaba Szepesv´ari, Arash Tavakoli, Cameron Upright, Yi Wan, Qing Yu, Yao- liang Yu, attendants of various seminars and webinars, in particular, a seminar at MIT on AlphaGo: Key Techniques and Applications, and an AI seminar at the University of Alberta on Deep Rein- forcement Learning: An Overview. Any remaining issues and errors are our own.
# REFERENCES
Abadi, M., Chu, A., Goodfellow, I., McMahan, H. B., Mironov, I., Talwar, K., and Zhang, L. (2016). Deep learning with differential privacy. In ACM Conference on Computer and Communications Security (ACM CCS).
Abbeel, P. and Ng, A. Y. (2004). Apprenticeship learning via inverse reinforcement learning. In the International Conference on Machine Learning (ICML). | 1701.07274#214 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 215 | Abbeel, P. and Ng, A. Y. (2004). Apprenticeship learning via inverse reinforcement learning. In the International Conference on Machine Learning (ICML).
Agrawal, P., Nair, A., Abbeel, P., Malik, J., and Levine, S. (2016). Learning to poke by poking: Experiential learning of intuitive physics. In the Annual Conference on Neural Information Pro- cessing Systems (NIPS).
Al-Shedivat, M., Bansal, T., Burda, Y., Sutskever, I., Mordatch, I., and Abbeel, P. (2017a). Con- tinuous Adaptation via Meta-Learning in Nonstationary and Competitive Environments. ArXiv e-prints.
Al-Shedivat, M., Dubey, A., and Xing, E. P. (2017b). Contextual Explanation Networks. ArXiv e-prints.
Albrechta, S. V. and Stone, P. (2018). Autonomous agents modelling other agents: A comprehensive survey and open problems. Artiï¬cial Intelligence. | 1701.07274#215 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 216 | Albrechta, S. V. and Stone, P. (2018). Autonomous agents modelling other agents: A comprehensive survey and open problems. Artiï¬cial Intelligence.
Alsheikh, M. A., Lin, S., Niyato, D., and Tan, H.-P. (2014). Machine learning in wireless sensor networks: Algorithms, strategies, and applications. IEEE Communications Surveys & Tutorials, 16(4):1996â2018.
Amin, K., Jiang, N., and Singh, S. (2017). Repeated inverse reinforcement learning. In the Annual Conference on Neural Information Processing Systems (NIPS).
Amodei, D., Olah, C., Steinhardt, J., Christiano, P., Schulman, J., and Man´e, D. (2016). Concrete Problems in AI Safety. ArXiv e-prints.
Anderson, H. S., Kharkar, A., Filar, B., and Roth, P. (2017). Evading machine learning malware detection. In Black Hat USA.
54
Anderson, R. N., Boulanger, A., Powell, W. B., and Scott, W. (2011). Adaptive stochastic control for the smart grid. Proceedings of the IEEE, 99(6):1098â1115. | 1701.07274#216 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 217 | Andreas, J., Klein, D., and Levine, S. (2017). Modular multitask reinforcement learning with policy sketches. In the International Conference on Machine Learning (ICML).
Andrychowicz, M., Denil, M., Colmenarejo, S. G., Hoffman, M. W., Pfau, D., Schaul, T., Shilling- ford, B., and de Freitas, N. (2016). Learning to learn by gradient descent by gradient descent. In the Annual Conference on Neural Information Processing Systems (NIPS).
Andrychowicz, M., Wolski, F., Ray, A., Schneider, J., Fong, R., Welinder, P., McGrew, B., Tobin, J., Abbeel, P., and Zaremba, W. (2017). Hindsight experience replay. In the Annual Conference on Neural Information Processing Systems (NIPS).
Anschel, O., Baram, N., and Shimkin, N. (2017). Averaged-DQN: Variance reduction and stabi- lization for deep reinforcement learning. In the International Conference on Machine Learning (ICML).
Argall, B. D., Chernova, S., Veloso, M., and Browning, B. (2009). A survey of robot learning from demonstration. Robotics and Autonomous Systems, 57(5):469â483. | 1701.07274#217 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 218 | Arjovsky, M., Chintala, S., and Bottou, L. (2017). Wasserstein GAN. ArXiv e-prints.
Artetxe, M., Labaka, G., Agirre, E., and Cho, K. (2017). Unsupervised Neural Machine Translation. ArXiv e-prints.
Arulkumaran, K., Deisenroth, M. P., Brundage, M., and Bharath, A. A. (2017). A Brief Survey of Deep Reinforcement Learning. ArXiv e-prints.
Asri, L. E., He, J., and Suleman, K. (2016). A sequence-to-sequence model for user simulation In Annual Meeting of the International Speech Communication in spoken dialogue systems. Association (INTERSPEECH).
Audiffren, J., Valko, M., Lazaric, A., and Ghavamzadeh, M. (2015). Maximum entropy semi- In the International Joint Conference on Artiï¬cial supervised inverse reinforcement learning. Intelligence (IJCAI).
Azar, M. G., Osband, I., and Munos, R. (2017). Minimax regret bounds for reinforcement learning. In the International Conference on Machine Learning (ICML). | 1701.07274#218 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 219 | Ba, J., Hinton, G. E., Mnih, V., Leibo, J. Z., and Ionescu, C. (2016). Using fast weights to attend to the recent past. In the Annual Conference on Neural Information Processing Systems (NIPS).
Ba, J., Mnih, V., and Kavukcuoglu, K. (2014). Multiple object recognition with visual attention. In the International Conference on Learning Representations (ICLR).
Babaeizadeh, M., Frosio, I., Tyree, S., Clemons, J., and Kautz, J. (2017). Reinforcement learn- ing through asynchronous advantage actor-critic on a gpu. In the International Conference on Learning Representations (ICLR).
Bacon, P.-L., Harb, J., and Precup, D. (2017). The option-critic architecture. In the AAAI Conference on Artiï¬cial Intelligence (AAAI).
Bahdanau, D., Brakel, P., Xu, K., Goyal, A., Lowe, R., Pineau, J., Courville, A., and Bengio, Y. (2017). An actor-critic algorithm for sequence prediction. In the International Conference on Learning Representations (ICLR). | 1701.07274#219 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 220 | Bahdanau, D., Cho, K., and Bengio, Y. (2015). Neural machine translation by jointly learning to align and translate. In the International Conference on Learning Representations (ICLR).
Bailis, P., Olukoton, K., Re, C., and Zaharia, M. (2017). Infrastructure for Usable Machine Learning: The Stanford DAWN Project. ArXiv e-prints.
Baird, L. (1995). Residual algorithms: Reinforcement learning with function approximation. In the International Conference on Machine Learning (ICML).
55
Baker, B., Gupta, O., Naik, N., and Raskar, R. (2017). Designing neural network architectures using reinforcement learning. In the International Conference on Learning Representations (ICLR).
Balle, B., Gomrokchi, M., and Precup, D. (2016). Differentially private policy evaluation. In the International Conference on Machine Learning (ICML).
Balog, M., Gaunt, A. L., Brockschmidt, M., Nowozin, S., and Tarlow, D. (2017). Deepcoder: Learning to write programs. In the International Conference on Learning Representations (ICLR). | 1701.07274#220 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 221 | Bansal, T., Pachocki, J., Sidor, S., Sutskever, I., and Mordatch, I. (2017). Emergent Complexity via Multi-Agent Competition. ArXiv e-prints.
Barreto, A., Munos, R., Schaul, T., and Silver, D. (2017). Successor features for transfer in rein- forcement learning. In the Annual Conference on Neural Information Processing Systems (NIPS).
Barto, A. G. and Mahadevan, S. (2003). Recent advances in hierarchical reinforcement learning. Discrete Event Dynamic Systems, 13(4):341â379.
Barto, A. G., Sutton, R. S., and Anderson, C. W. (1983). Neuronlike elements that can solve difï¬cult learning control problems. IEEE Transactions on Systems, Man, and Cybernetics, 13:835â846.
Battaglia, P. W., Pascanu, R., Lai, M., Rezende, D., and Kavukcuoglu, K. (2016). Interaction net- In the Annual Conference on Neural works for learning about objects, relations and physics. Information Processing Systems (NIPS). | 1701.07274#221 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 222 | Bazzan, A. L. and Kl¨ugl, F. (2014). Introduction to Intelligent Systems in Trafï¬c and Transportation. Morgan & Claypool.
Beattie, C., Leibo, J. Z., Teplyashin, D., Ward, T., Wainwright, M., K¨uttler, H., Lefrancq, A., Green, S., Vald´es, V., Sadik, A., Schrittwieser, J., Anderson, K., York, S., Cant, M., Cain, A., Bolton, A., Gaffney, S., King, H., Hassabis, D., Legg, S., and Petersen, S. (2016). DeepMind Lab. ArXiv e-prints.
Bellemare, M. G., Dabney, W., and Munos, R. (2017). A distributional perspective on reinforcement learning. In the International Conference on Machine Learning (ICML).
Bellemare, M. G., Danihelka, I., Dabney, W., Mohamed, S., Lakshminarayanan, B., Hoyer, S., and Munos, R. (2017). The Cramer Distance as a Solution to Biased Wasserstein Gradients. ArXiv e-prints. | 1701.07274#222 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 223 | Bellemare, M. G., Naddaf, Y., Veness, J., and Bowling, M. (2013). The arcade learning environment: An evaluation platform for general agents. Journal of Artiï¬cial Intelligence Research, 47:253â 279.
Bellemare, M. G., Schaul, T., Srinivasan, S., Saxton, D., Ostrovski, G., and Munos, R. (2016). Unifying count-based exploration and intrinsic motivation. In the Annual Conference on Neural Information Processing Systems (NIPS).
Bello, I., Pham, H., Le, Q. V., Norouzi, M., and Bengio, S. (2016). Neural Combinatorial Optimiza- tion with Reinforcement Learning. ArXiv e-prints.
Bello, I., Zoph, B., Vasudevan, V., and Le, Q. V. (2017). Neural optimizer search with reinforcement learning. In the International Conference on Machine Learning (ICML).
Bengio, Y. (2009). Learning deep architectures for ai. Foundations and trends®in Machine Learn- ing, 2(1):1-127.
Bengio, Y. (2017). The Consciousness Prior. ArXiv e-prints. | 1701.07274#223 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 224 | Bengio, Y. (2017). The Consciousness Prior. ArXiv e-prints.
Bengio, Y., Louradour, J., Collobert, R., and Weston, J. (2009). Curriculum learning. In the Inter- national Conference on Machine Learning (ICML).
Berkenkamp, F., Turchetta, M., Schoellig, A. P., and Krause, A. (2017). Safe model-based rein- forcement learning with stability guarantees. In the Annual Conference on Neural Information Processing Systems (NIPS).
56
Bernhard Wymann, E. E., Guionneau, C., Dimitrakakis, C., and R´emi Coulom, A. S. (2014). TORCS, The Open Racing Car Simulator. âhttp://www.torcs.orgâ.
Berthelot, D., Schumm, T., and Metz, L. (2017). BEGAN: Boundary Equilibrium Generative Ad- versarial Networks. ArXiv e-prints.
Bertsekas, D. P. (2012). Dynamic programming and optimal control (Vol. II, 4th Edition: Approxi- mate Dynamic Programming). Athena Scientiï¬c, Massachusetts, USA. | 1701.07274#224 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 225 | Bertsekas, D. P. and Tsitsiklis, J. N. (1996). Neuro-Dynamic Programming. Athena Scientiï¬c.
Bhatti, S., Desmaison, A., Miksik, O., Nardelli, N., Siddharth, N., and Torr, P. H. S. (2016). Playing Doom with SLAM-Augmented Deep Reinforcement Learning. ArXiv e-prints.
Bishop, C. (2011). Pattern Recognition and Machine Learning. Springer.
Blei, D. M. and Smyth, P. (2017). Science and data science. PNAS, 114(33):8689â8692.
Bohg, J., Hausman, K., Sankaran, B., Brock, O., Kragic, D., Schaal, S., and Sukhatme, G. S. (2017). Interactive perception: Leveraging action in perception and perception in action. IEEE Transac- tions on Robotics, 33(6):1273â1291. | 1701.07274#225 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 226 | Bojarski, M., Testa, D. D., Dworakowski, D., Firner, B., Flepp, B., Goyal, P., Jackel, L. D., Monfort, M., Muller, U., Zhang, J., Zhang, X., Zhao, J., and Zieba, K. (2016). End to End Learning for Self-Driving Cars. ArXiv e-prints.
Bojarski, M., Yeres, P., Choromanska, A., Choromanski, K., Firner, B., Jackel, L., and Muller, U. (2017). Explaining How a Deep Neural Network Trained with End-to-End Learning Steers a Car. ArXiv e-prints.
Bordes, A., Boureau, Y.-L., and Weston, J. (2017). Learning end-to-end goal-oriented dialog. In the International Conference on Learning Representations (ICLR). | 1701.07274#226 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 227 | Botvinick, M., Barrett, D. G. T., Battaglia, P., de Freitas, N., Kumaran, D., Leibo, J. Z., Lillicrap, T., Modayil, J., Mohamed, S., Rabinowitz, N. C., Rezende, D. J., Santoro, A., Schaul, T., Sum- merï¬eld, C., Wayne, G., Weber, T., Wierstra, D., Legg, S., and Hassabis, D. (2017). Building machines that learn and think for themselves. Behavioral and Brain Sciences, 40.
Bousmalis, K., Irpan, A., Wohlhart, P., Bai, Y., Kelcey, M., Kalakrishnan, M., Downs, L., Ibarz, J., Pastor, P., Konolige, K., Levine, S., and Vanhoucke, V. (2017). Using Simulation and Domain Adaptation to Improve Efï¬ciency of Deep Robotic Grasping. ArXiv e-prints.
Bowling, M., Burch, N., Johanson, M., and Tammelin, O. (2015). Heads-up limit holdâem poker is solved. Science, 347(6218):145â149. | 1701.07274#227 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 228 | Boyd, S. and Vandenberghe, L. (2004). Convex Optimization. Cambridge University Press.
Bradtke, S. J. and Barto, A. G. (1996). Linear least-squares algorithms for temporal difference learning. Machine Learning, 22(1-3):33â57.
Brandt, M. W., Goyal, A., Santa-Clara, P., and Stroud, J. R. (2005). A simulation approach to dynamic portfolio choice with an application to learning about return predictability. The Review of Financial Studies, 18(3):831â873.
Briot, J.-P., Hadjeres, G., and Pachet, F. (2017). Deep Learning Techniques for Music Generation - A Survey. ArXiv e-prints.
Browne, C., Powley, E., Whitehouse, D., Lucas, S., Cowling, P. I., Rohlfshagen, P., Tavener, S., Perez, D., Samothrakis, S., and Colton, S. (2012). A survey of Monte Carlo tree search methods. IEEE Transactions on Computational Intelligence and AI in Games, 4(1):1â43. | 1701.07274#228 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 229 | Brunner, G., Richter, O., Wang, Y., and Wattenhofer, R. (2018). Teaching a machine to read maps with deep reinforcement learning. In the AAAI Conference on Artiï¬cial Intelligence (AAAI).
57
Busoniu, L., Babuska, R., and Schutter, B. D. (2008). A comprehensive survey of multiagent rein- forcement learning. IEEE Transactions on Systems, Man, and Cybernetics - Part C: Applications and Reviews, 38(2).
Cai, J., Shin, R., and Song, D. (2017). Making neural programming architectures generalize via recursion. In the International Conference on Learning Representations (ICLR).
Caicedo, J. C. and Lazebnik, S. (2015). Active object localization with deep reinforcement learning. In the IEEE International Conference on Computer Vision (ICCV).
Cao, Q., Lin, L., Shi, Y., Liang, X., and Li, G. (2017). Attention-aware face hallucination via deep reinforcement learning. In the IEEE Conference on Computer Vision and Pattern Recognition (CVPR).
Carleo, G. and Troyer, M. (2017). Solving the quantum many-body problem with artiï¬cial neural networks. Science, 355(6325):602â606. | 1701.07274#229 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 230 | Carlini, N. and Wagner, D. (2017a). Adversarial examples are not easily detected: Bypassing ten detection methods. In ACM CCS 2017 Workshop on Artiï¬cial Intelligence and Security.
Carlini, N. and Wagner, D. (2017b). Towards evaluating the robustness of neural networks. In IEEE Symposium on Security and Privacy.
Celikyilmaz, A., Deng, L., Li, L., and Wang, C. (2017). Scaffolding Networks: Incremental Learn- ing and Teaching Through Questioning. ArXiv e-prints.
Chakraborty, B. and Murphy, S. A. (2014). Dynamic treatment regimes. Annual Review of Statistics and Its Application, 1:447â464.
Chebotar, Y., Hausman, K., Zhang, M., Sukhatme, G., Schaal, S., and Levine, S. (2017). Com- bining model-based and model-free updates for trajectory-centric reinforcement learning. In the International Conference on Machine Learning (ICML).
Chebotar, Y., Kalakrishnan, M., Yahya, A., Li, A., Schaal, S., and Levine, S. (2016). Path integral guided policy search. ArXiv e-prints. | 1701.07274#230 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 231 | Chen, J., Huang, P.-S., He, X., Gao, J., and Deng, L. (2016). Unsupervised Learning of Predictors from Unpaired Input-Output Samples. ArXiv e-prints.
Chen, X., Duan, Y., Houthooft, R., Schulman, J., Sutskever, I., and Abbeel, P. (2016a). InfoGAN: Interpretable representation learning by information maximizing generative adversarial nets. In the Annual Conference on Neural Information Processing Systems (NIPS).
Chen, Y.-N., Hakkani-Tur, D., Tur, G., Celikyilmaz, A., Gao, J., and Deng, L. (2016b). Knowledge as a Teacher: Knowledge-Guided Structural Attention Networks. ArXiv e-prints.
Chen, Y.-N. V., Hakkani-T¨ur, D., Tur, G., Gao, J., and Deng, L. (2016c). End-to-end memory In Annual networks with knowledge carryover for multi-turn spoken language understanding. Meeting of the International Speech Communication Association (INTERSPEECH).
Chen, Z. and Liu, B. (2016). Lifelong Machine Learning. Morgan & Claypool Publishers. | 1701.07274#231 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 232 | Chen, Z. and Liu, B. (2016). Lifelong Machine Learning. Morgan & Claypool Publishers.
Chen, Z. and Yi, D. (2017). The Game Imitation: Deep Supervised Convolutional Networks for Quick Video Game AI. ArXiv e-prints.
Cheng, Y., Xu, W., He, Z., He, W., Wu, H., Sun, M., and Liu, Y. (2016). Semi-supervised learning for neural machine translation. In the Association for Computational Linguistics annual meeting (ACL).
Cho, K. (2015). Natural Language Understanding with Distributed Representation. ArXiv e-prints.
Cho, K., van Merrienboer, B., Gulcehre, C., Bougares, F., Schwenk, H., and Bengio, Y. (2014). Learning phrase representations using RNN encoder-decoder for statistical machine translation. In Conference on Empirical Methods in Natural Language Processing (EMNLP).
58
Choi, E., Hewlett, D., Polosukhin, I., Lacoste, A., Uszkoreit, J., and Berant, J. (2017). Coarse-to-ï¬ne question answering for long documents. In the Association for Computational Linguistics annual meeting (ACL). | 1701.07274#232 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 233 | Christiano, P., Leike, J., Brown, T. B., Martic, M., Legg, S., and Amodei, D. (2017). Deep rein- forcement learning from human preferences. In the Annual Conference on Neural Information Processing Systems (NIPS).
Chung, J., Gulcehre, C., Cho, K., and Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. In NIPS 2014 Deep Learning and Representation Learn- ing Workshop.
Crawford, D., Levit, A., Ghadermarzy, N., Oberoi, J. S., and Ronagh, P. (2016). Reinforcement Learning Using Quantum Boltzmann Machines. ArXiv e-prints.
Czarnecki, W. M., ´Swirszcz, G., Jaderberg, M., Osindero, S., Vinyals, O., and Kavukcuoglu, K. (2017). Understanding Synthetic Gradients and Decoupled Neural Interfaces. ArXiv e-prints.
Dai, H., Khalil, E. B., Zhang, Y., Dilkina, B., and Song, L. (2017). Learning combinatorial opti- mization algorithms over graphs. In the Annual Conference on Neural Information Processing Systems (NIPS). | 1701.07274#233 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 234 | Dai, Z., Yang, Z., Yang, F., Cohen, W. W., and Salakhutdinov, R. (2017). Good Semi-supervised Learning that Requires a Bad GAN. ArXiv e-prints.
Daniely, A., Frostig, R., and Singer, Y. (2016). Toward deeper understanding of neural networks: The power of initialization and a dual view on expressivity. In the Annual Conference on Neural Information Processing Systems (NIPS).
Danihelka, I., Wayne, G., Uria, B., Kalchbrenner, N., and Graves, A. (2016). Associative long short-term memory. In the International Conference on Machine Learning (ICML).
Das, A., Kottur, S., Moura, J. M. F., Lee, S., and Batra, D. (2017). Learning cooperative visual dialog In the IEEE International Conference on Computer agents with deep reinforcement learning. Vision (ICCV).
De Asis, K., Hernandez-Garcia, J. F., Zacharias Holland, G., and Sutton, R. S. (2018). Multi-step reinforcement learning: A unifying algorithm. In the AAAI Conference on Artiï¬cial Intelligence (AAAI). | 1701.07274#234 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 235 | Deisenroth, M. P., Neumann, G., and Peters, J. (2013). A survey on policy search for robotics. Foundations and Trend in Robotics, 2:1â142.
Deisenroth, M. P. and Rasmussen, C. E. (2011). PILCO: A model-based and data-efï¬cient approach to policy search. In the International Conference on Machine Learning (ICML).
Delle Fave, F. M., Jiang, A. X., Yin, Z., Zhang, C., Tambe, M., Kraus, S., and Sullivan, J. P. (2014). Game-theoretic security patrolling with dynamic execution uncertainty and a case study on a real transit system. 50:321â367.
Deng, L. (2017). Three generations of talk at https://www.slideshare.net/AIFrontiers/ spoken dialogue systems (bots), AI li-deng-three-generations-of-spoken-dialogue-systems-bots. Frontiers Conference.
Deng, L. and Dong, Y. (2014). Deep Learning: Methods and Applications. Now Publishers Inc.
Deng, L. and Li, X. (2013). Machine learning paradigms for speech recognition: An overview. IEEE Transactions on Audio, Speech, and Language Processing, 21(5):1060â1089. | 1701.07274#235 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 236 | Deng, L. and Liu, Y. (2017). Deep Learning in Natural Language Processing (edited book, sched- uled August 2017). Springer.
Deng, Y., Bao, F., Kong, Y., Ren, Z., and Dai, Q. (2016). Deep direct reinforcement learning for ï¬nancial signal representation and trading. IEEE Transactions on Neural Networks and Learning Systems.
59
Denil, M., Agrawal, P., Kulkarni, T. D., Erez, T., Battaglia, P., and de Freitas, N. (2017). Learning to perform physics experiments via deep reinforcement learning. In the International Conference on Learning Representations (ICLR).
Denil, M., G´omez Colmenarejo, S., Cabi, S., Saxton, D., and de Freitas, N. (2017). Programmable Agents. ArXiv e-prints.
Devrim Kaba, M., Gokhan Uzunbas, M., and Nam Lim, S. (2017). A reinforcement learning ap- proach to the view planning problem. In the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). | 1701.07274#236 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 237 | Dhingra, B., Li, L., Li, X., Gao, J., Chen, Y.-N., Ahmed, F., and Deng, L. (2017). End-to-end reinforcement learning of dialogue agents for information access. In the Association for Compu- tational Linguistics annual meeting (ACL).
Diederik P Kingma, M. W. (2014). Auto-encoding variational bayes. In the International Conference on Learning Representations (ICLR).
Dietterich, T. G. (2000). Hierarchical reinforcement learning with the MAXQ value function de- composition. Journal of Artiï¬cial Intelligence Research, 13(1):227â303.
Domingos, P. (2012). A few useful things to know about machine learning. Communications of the ACM, 55(10):78â87.
Dong, D., Wu, H., He, W., Yu, D., and Wang, H. (2015). Multi-task learning for multiple language translation. In the Association for Computational Linguistics annual meeting (ACL).
Dong, Y., Su, H., Zhu, J., and Zhang, B. (2017). Improving interpretability of deep neural networks with semantic information. In the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). | 1701.07274#237 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 238 | Doshi-Velez, F. and Kim, B. (2017). Towards A Rigorous Science of Interpretable Machine Learn- ing. ArXiv e-prints.
Dosovitskiy, A. and Koltun, V. (2017). Learning to act by predicting the future. In the International Conference on Learning Representations (ICLR).
Downey, C., Hefny, A., Li, B., Boots, B., and Gordon, G. (2017). Predictive state recurrent neural networks. In the Annual Conference on Neural Information Processing Systems (NIPS).
Du, S. S., Chen, J., Li, L., Xiao, L., and Zhou, D. (2017). Stochastic variance reduction methods for policy evaluation. In the International Conference on Machine Learning (ICML).
Duan, Y., Andrychowicz, M., Stadie, B. C., Ho, J., Schneider, J., Sutskever, I., Abbeel, P., and Zaremba, W. (2017). One-shot imitation learning. In the Annual Conference on Neural Informa- tion Processing Systems (NIPS).
Duan, Y., Chen, X., Houthooft, R., Schulman, J., and Abbeel, P. (2016). Benchmarking deep rein- forcement learning for continuous control. In the International Conference on Machine Learning (ICML). | 1701.07274#238 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 239 | Duan, Y., Schulman, J., Chen, X., Bartlett, P. L., Sutskever, I., and Abbeel, P. (2016). RL2: Fast Reinforcement Learning via Slow Reinforcement Learning. ArXiv e-prints.
Dulac-Arnold, G., Evans, R., van Hasselt, H., Sunehag, P., Lillicrap, T., Hunt, J., Mann, T., Weber, T., Degris, T., and Coppin, B. (2015). Deep Reinforcement Learning in Large Discrete Action Spaces. ArXiv e-prints.
El-Tantawy, S., Abdulhai, B., and Abdelgawad, H. (2013). Multiagent reinforcement learning for integrated network of adaptive trafï¬c signal controllers (marlin-atsc): methodology and large- scale application on downtown toronto. IEEE Transactions on Intelligent Transportation Systems, 14(3):1140â1150.
Eric, M. and Manning, C. D. (2017). A Copy-Augmented Sequence-to-Sequence Architecture Gives Good Performance on Task-Oriented Dialogue. ArXiv e-prints.
60 | 1701.07274#239 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 240 | 60
Ernst, D., Geurts, P., and Wehenkel, L. (2005). Tree-based batch mode reinforcement learning. The Journal of Machine Learning Research, 6:503â556.
Eslami, S. M. A., Heess, N., Weber, T., Tassa, Y., Szepesv´ari, D., Kavukcuoglu, K., and Hinton, In the G. E. (2016). Attend, infer, repeat: Fast scene understanding with generative models. Annual Conference on Neural Information Processing Systems (NIPS).
Evans, R. and Grefenstette, E. (2017). Learning Explanatory Rules from Noisy Data. ArXiv e-prints.
Evtimov, I., Eykholt, K., Fernandes, E., Kohno, T., Li, B., Prakash, A., Rahmati, A., and Song, D. (2017). Robust Physical-World Attacks on Deep Learning Models. ArXiv e-prints.
Fang, M., Li, Y., and Cohn, T. (2017). Learning how to active learn: A deep reinforcement learning approach. In Conference on Empirical Methods in Natural Language Processing (EMNLP). | 1701.07274#240 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 241 | Fang, X., Misra, S., Xue, G., and Yang, D. (2012). Smart grid - the new and improved power grid: A survey. IEEE Communications Surveys Tutorials, 14(4):944â980.
Fatemi, M., Asri, L. E., Schulz, H., He, J., and Suleman, K. (2016). Policy networks with two- stage training for dialogue systems. In the Annual SIGdial Meeting on Discourse and Dialogue (SIGDIAL).
Feng, J. and Zhou, Z.-H. (2017). AutoEncoder by Forest. ArXiv e-prints.
Fernando, C., Banarse, D., Blundell, C., Zwols, Y., Ha, D., Rusu, A. A., Pritzel, A., and Wierstra, D. (2017). PathNet: Evolution Channels Gradient Descent in Super Neural Networks. ArXiv e-prints.
Finn, C., Christiano, P., Abbeel, P., and Levine, S. (2016a). A connection between GANs, inverse reinforcement learning, and energy-based models. In NIPS 2016 Workshop on Adversarial Train- ing. | 1701.07274#241 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 242 | Finn, C. and Levine, S. (2016). Deep visual foresight for planning robot motion. In IEEE Interna- tional Conference on Robotics and Automation (ICRA).
Finn, C., Levine, S., and Abbeel, P. (2016b). Guided cost learning: Deep inverse optimal control via policy optimization. In the International Conference on Machine Learning (ICML).
Finn, C., Yu, T., Fu, J., Abbeel, P., and Levine, S. (2017). Generalizing skills with semi-supervised reinforcement learning. In the International Conference on Learning Representations (ICLR).
Firoiu, V., Whitney, W. F., and Tenenbaum, J. B. (2017). Beating the Worldâs Best at Super Smash Bros. with Deep Reinforcement Learning. ArXiv e-prints.
Florensa, C., Duan, Y., and Abbeel, P. (2017). Stochastic neural networks for hierarchical reinforce- ment learning. In the International Conference on Learning Representations (ICLR).
Foerster, J., Assael, Y. M., de Freitas, N., and Whiteson, S. (2016). Learning to communicate with deep multi-agent reinforcement learning. In the Annual Conference on Neural Information Processing Systems (NIPS). | 1701.07274#242 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 243 | Foerster, J., Farquhar, G., Afouras, T., Nardelli, N., and Whiteson, S. (2018). Counterfactual multi- agent policy gradients. In the AAAI Conference on Artiï¬cial Intelligence (AAAI).
Foerster, J., Nardelli, N., Farquhar, G., Torr, P. H. S., Kohli, P., and Whiteson, S. (2017). Stabilising experience replay for deep multi-agent reinforcement learning. In the International Conference on Machine Learning (ICML).
Foerster, J. N., Chen, R. Y., Al-Shedivat, M., Whiteson, S., Abbeel, P., and Mordatch, I. (2017). Learning with Opponent-Learning Awareness. ArXiv e-prints.
Fortunato, M., Gheshlaghi Azar, M., Piot, B., Menick, J., Osband, I., Graves, A., Mnih, V., Munos, R., Hassabis, D., Pietquin, O., Blundell, C., and Legg, S. (2017). Noisy Networks for Exploration. ArXiv e-prints.
61 | 1701.07274#243 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 244 | 61
Fu, J., Co-Reyes, J. D., and Levine, S. (2017). Ex2: Exploration with exemplar models for deep In the Annual Conference on Neural Information Processing Systems reinforcement learning. (NIPS).
Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., and Lempitsky, V. (2016). Domain-adversarial training of neural networks. Journal of Machine Learning Research, 17(59):1â35.
Garc`ıa, J. and Fern`andez, F. (2015). A comprehensive survey on safe reinforcement learning. The Journal of Machine Learning Research, 16:1437â1480.
Gavrilovska, L., Atanasovski, V., Macaluso, I., and DaSilva, L. A. (2013). Learning and reasoning in cognitive radio networks. IEEE Communications Surveys Tutorials, 15(4):1761â1777.
Gehring, J., Auli, M., Grangier, D., Yarats, D., and Dauphin, Y. N. (2017). Convolutional Sequence to Sequence Learning. ArXiv e-prints. | 1701.07274#244 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 245 | Gelly, S., Schoenauer, M., Sebag, M., Teytaud, O., Kocsis, L., Silver, D., and Szepesv´ari, C. (2012). The grand challenge of computer go: Monte carlo tree search and extensions. Communications of the ACM, 55(3):106â113.
Gelly, S. and Silver, D. (2007). Combining online and ofï¬ine knowledge in uct. In the International Conference on Machine Learning (ICML).
George, D., Lehrach, W., Kansky, K., L´azaro-Gredilla, M., Laan, C., Marthi, B., Lou, X., Meng, Z., Liu, Y., Wang, H., Lavin, A., and Phoenix, D. S. (2017). A generative vision model that trains with high data efï¬ciency and breaks text-based CAPTCHAs. Science.
Geramifard, A., Dann, C., Klein, R. H., Dabney, W., and How, J. P. (2015). Rlpy: A value-function- based reinforcement learning framework for education and research. Journal of Machine Learning Research, 16:1573â1578. | 1701.07274#245 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 246 | Geramifard, A., Walsh, T. J., Tellex, S., Chowdhary, G., Roy, N., and How, J. P. (2013). A tutorial on linear function approximators for dynamic programming and reinforcement learning. Foundations and Trends in Machine Learning, 6(4):375â451.
Ghavamzadeh, M., Mahadevan, S., and Makar, R. (2006). Hierarchical multi-agent reinforcement learning. Autonomous Agents and Multi-Agent Systems, 13(2):197â229.
Ghavamzadeh, M., Mannor, S., Pineau, J., and Tamar, A. (2015). Bayesian reinforcement learning: a survey. Foundations and Trends in Machine Learning, 8(5-6):359â483.
Girshick, R. (2015). Fast R-CNN. In the IEEE International Conference on Computer Vision (ICCV).
Glavic, M., Fonteneau, R., and Ernst, D. (2017). Reinforcement learning for electric power system decision and control: Past considerations and perspectives. In The 20th World Congress of the International Federation of Automatic Control.
Goldberg, Y. (2017). Neural Network Methods for Natural Language Processing. Morgan & Clay- pool Publishers. | 1701.07274#246 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 247 | Goldberg, Y. (2017). Neural Network Methods for Natural Language Processing. Morgan & Clay- pool Publishers.
Goldberg, Y. and Kosorok, M. R. (2012). Q-learning with censored data. Annals of Statistics, 40(1):529â560.
Goodfellow, I. (2017). NIPS 2016 Tutorial: Generative Adversarial Networks. ArXiv e-prints.
Goodfellow, I., Bengio, Y., and Courville, A. (2016). Deep Learning. MIT Press.
Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., , and Bengio, Y. (2014). Generative adversarial nets. In the Annual Conference on Neural Infor- mation Processing Systems (NIPS), page 2672?2680.
Graves, A., Bellemare, M. G., Menick, J., Munos, R., and Kavukcuoglu, K. (2017). Automated Curriculum Learning for Neural Networks. ArXiv e-prints.
62
Graves, A., Wayne, G., and Danihelka, I. (2014). Neural Turing Machines. ArXiv e-prints. | 1701.07274#247 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 248 | 62
Graves, A., Wayne, G., and Danihelka, I. (2014). Neural Turing Machines. ArXiv e-prints.
Graves, A., Wayne, G., Reynolds, M., Harley, T., Danihelka, I., Grabska-Barwi´nska, A., Col- menarejo, S. G., Grefenstette, E., Ramalho, T., Agapiou, J., nech Badia, A. P., Hermann, K. M., Zwols, Y., Ostrovski, G., Cain, A., King, H., Summerï¬eld, C., Blunsom, P., Kavukcuoglu, K., and Hassabis, D. (2016). Hybrid computing using a neural network with dynamic external memory. Nature, 538:471â476.
Gregor, K., Danihelka, I., Graves, A., Rezende, D., and Wierstra, D. (2015). Draw: A recurrent In the International Conference on Machine Learning neural network for image generation. (ICML). | 1701.07274#248 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 249 | Grondman, I., Busoniu, L., Lopes, G. A., and BabuËska, R. (2012). A survey of actor-critic rein- forcement learning: Standard and natural policy gradients. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), 42(6):1291â1307.
Gruslys, A., Gheshlaghi Azar, M., Bellemare, M. G., and Munos, R. (2017). The Reactor: A Sample-Efï¬cient Actor-Critic Architecture. ArXiv e-prints.
Gu, S., Holly, E., Lillicrap, T., and Levine, S. (2016a). Deep reinforcement learning for robotic manipulation with asynchronous off-policy updates. ArXiv e-prints.
Gu, S., Lillicrap, T., Ghahramani, Z., Turner, R. E., and Levine, S. (2017). Q-Prop: Sample- efï¬cient policy gradient with an off-policy critic. In the International Conference on Learning Representations (ICLR). | 1701.07274#249 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 250 | Gu, S., Lillicrap, T., Ghahramani, Z., Turner, R. E., Sch¨olkopf, B., and Levine, S. (2017). Interpo- lated policy gradient: Merging on-policy and off-policy gradient estimation for deep reinforce- ment learning. In the Annual Conference on Neural Information Processing Systems (NIPS).
Gu, S., Lillicrap, T., Sutskever, I., and Levine, S. (2016b). Continuous deep Q-learning with model- based acceleration. In the International Conference on Machine Learning (ICML).
Gulcehre, C., Chandar, S., Cho, K., and Bengio, Y. (2016). Dynamic Neural Turing Machine with Soft and Hard Addressing Schemes. ArXiv e-prints.
Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., and Courville, A. (2017). Improved training of wasserstein gans. In the Annual Conference on Neural Information Processing Systems (NIPS).
Gupta, A., Devin, C., Liu, Y., Abbeel, P., and Levine, S. (2017a). Learning invariant feature spaces to transfer skills with reinforcement learning. In the International Conference on Learning Rep- resentations (ICLR). | 1701.07274#250 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
1701.07274 | 251 | Gupta, S., Davidson, J., Levine, S., Sukthankar, R., and Malik, J. (2017b). Cognitive mapping In the IEEE Conference on Computer Vision and Pattern and planning for visual navigation. Recognition (CVPR).
Guu, K., Pasupat, P., Liu, E. Z., and Liang, P. (2017). From language to programs: Bridging reinforcement learning and maximum marginal likelihood. In the Association for Computational Linguistics annual meeting (ACL).
Haarnoja, T., Tang, H., Abbeel, P., and Levine, S. (2017). Reinforcement learning with deep energy- based policies. In the International Conference on Machine Learning (ICML).
Hadï¬eld-Menell, D., Dragan, A., Abbeel, P., and Russell, S. (2016). Cooperative inverse reinforce- ment learning. In the Annual Conference on Neural Information Processing Systems (NIPS).
Hadï¬eld-Menell, D., Milli, S., Abbeel, P., Russell, S., and Dragan, A. (2017). Inverse reward design. In the Annual Conference on Neural Information Processing Systems (NIPS). | 1701.07274#251 | Deep Reinforcement Learning: An Overview | We give an overview of recent exciting achievements of deep reinforcement
learning (RL). We discuss six core elements, six important mechanisms, and
twelve applications. We start with background of machine learning, deep
learning and reinforcement learning. Next we discuss core RL elements,
including value function, in particular, Deep Q-Network (DQN), policy, reward,
model, planning, and exploration. After that, we discuss important mechanisms
for RL, including attention and memory, unsupervised learning, transfer
learning, multi-agent RL, hierarchical RL, and learning to learn. Then we
discuss various applications of RL, including games, in particular, AlphaGo,
robotics, natural language processing, including dialogue systems, machine
translation, and text generation, computer vision, neural architecture design,
business management, finance, healthcare, Industry 4.0, smart grid, intelligent
transportation systems, and computer systems. We mention topics not reviewed
yet, and list a collection of RL resources. After presenting a brief summary,
we close with discussions.
Please see Deep Reinforcement Learning, arXiv:1810.06339, for a significant
update. | http://arxiv.org/pdf/1701.07274 | Yuxi Li | cs.LG | Please see Deep Reinforcement Learning, arXiv:1810.06339, for a
significant update | null | cs.LG | 20170125 | 20181126 | [] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.