deep-reinforcement-learning-records

Explicitly show the relationships between various techniques of deep reinforcement learning methods. Dedicated for learning and researching on DRL.

Recommendations and suggestions are welcome.

Value based methods

• Deep Q-Networks
• Double DQN
• Prioritized Experience
• Dueling Network Architecture
• Learning from Demonstrations for Real World Reinforcement Learning

Policy gradient methods

• Proximal Policy Optimization Algorithms 20 July 2017
• Emergence of Locomotion Behaviours in Rich Environments 7 July 2017
• Equivalence Between Policy Gradients and Soft Q-Learning
• Trust Region Policy Optimization
• Reinforcement Learning with Deep Energy-Based Policies
• Q-PROP: SAMPLE-EFFICIENT POLICY GRADIENT WITH AN OFF-POLICY CRITIC
• Interpolated Policy Gradient: Merging On-Policy and Off-Policy Gradient Estimation for Deep Reinforcement Learning 1 Jun 2017

Explorations in DRL

• Noisy Networks for Exploration 30 Jun 2017 implementation
• Count-Based Exploration in Feature Space for Reinforcement Learning 25 Jun 2017
• Count-Based Exploration with Neural Density Models 14 Jun 2017
• UCB and InfoGain Exploration via Q-Ensembles 11 Jun 2017
• Minimax Regret Bounds for Reinforcement Learning 16 Mar 2017
• Incentivizing Exploration In Reinforcement Learning With Deep Predictive Models
• EX2: Exploration with Exemplar Models for Deep Reinforcement Learning

Actor-Critic methods

• The Reactor: A Sample-Efficient Actor-Critic Architecture 15 Apr 2017
• SAMPLE EFFICIENT ACTOR-CRITIC WITH EXPERIENCE REPLAY
• REINFORCEMENT LEARNING WITH UNSUPERVISED AUXILIARY TASKS
• Continuous control with deep reinforcement learning

Model-based methods

• Learning model-based planning from scratch 19 July 2017

Model-free + Model-based

• Imagination-Augmented Agents for Deep Reinforcement Learning 19 July 2017

Option

• A Laplacian Framework for Option Discovery in Reinforcement Learning 16 Jun 2017

Connection with other methods

• Robust Imitation of Diverse Behaviors
• Learning human behaviors from motion capture by adversarial imitation
• Connecting Generative Adversarial Networks and Actor-Critic Methods

Connecting value and policy methods

• Bridging the Gap Between Value and Policy Based Reinforcement Learning
• Policy gradient and Q-learning

Reward design

• Reinforcement Learning with Corrupted Reward Channel 23 May 2017

Unifying

• Multi-step Reinforcement Learning: A Unifying Algorithm

Faster DRL

• Neural Episodic Control

Apply RL to other domains

• TUNING RECURRENT NEURAL NETWORKS WITH REINFORCEMENT LEARNING

Multiagent Settings

• Counterfactual Multi-Agent Policy Gradients
• Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments 7 Jun 2017
• Multiagent Bidirectionally-Coordinated Nets for Learning to Play StarCraft Combat Games 29 Mar 2017

New design

• Reverse Curriculum Generation for Reinforcement Learning
• Trial without Error: Towards Safe Reinforcement Learning via Human Intervention
• Learning to Design Games: Strategic Environments in Deep Reinforcement Learning 5 July 2017

Multitask

• Distral: Robust Multitask Reinforcement Learning 13 July 2017