Deep Reinforcement and InfoMax Learning

Abstract

We begin with the hypothesis that a model-free agent whose representations are predictive of properties of future states (beyond expected rewards) will be more capable of solving and adapting to new RL problems. To test that hypothesis, we introduce an objective based on Deep InfoMax (DIM) which trains the agent to predict the future by maximizing the mutual information between its internal representation of successive timesteps. We test our approach in several synthetic settings, where it successfully learns representations that are predictive of the future. Finally, we augment C51, a strong RL baseline, with our temporal DIM objective and demonstrate improved performance on a continual learning task and on the recently introduced Procgen environment.

arXiv commentNeurIPS 2020

References in corpus (23)

• A Simple Framework for Contrastive Learning of Visual Representations
• Representation Learning with Contrastive Predictive Coding
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• Playing Atari with Deep Reinforcement Learning
• Learning deep representations by mutual information estimation and maximization
• Momentum Contrast for Unsupervised Visual Representation Learning
• Data-Efficient Image Recognition with Contrastive Predictive Coding
• Contrastive Multiview Coding
• MINE: Mutual Information Neural Estimation
• Learning Latent Dynamics for Planning from Pixels
• Variational Information Maximisation for Intrinsically Motivated Reinforcement Learning
• Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICA
• Temporal Difference Models: Model-Free Deep RL for Model-Based Control
• Dream to Control: Learning Behaviors by Latent Imagination
• Pseudo-Rehearsal: Achieving Deep Reinforcement Learning without Catastrophic Forgetting
• Generalization and Regularization in DQN
• Sharing Knowledge in Multi-Task Deep Reinforcement Learning
• Unicorn: Continual Learning with a Universal, Off-policy Agent
• EMI: Exploration with Mutual Information
• Continual Reinforcement Learning with Complex Synapses
• A Theoretical Analysis of Catastrophic Forgetting through the NTK Overlap Matrix
• Batch Stationary Distribution Estimation
• Representation of Reinforcement Learning Policies in Reproducing Kernel Hilbert Spaces

Cited by in corpus (9)

• Decoupling Representation Learning from Reinforcement Learning
• Data-Efficient Reinforcement Learning with Self-Predictive Representations
• Return-Based Contrastive Representation Learning for Reinforcement Learning
• Multimodal Information Bottleneck for Deep Reinforcement Learning with Multiple Sensors
• Which Mutual-Information Representation Learning Objectives are Sufficient for Control?
• Cross-Trajectory Representation Learning for Zero-Shot Generalization in RL
• Domain Adversarial Reinforcement Learning
• DRIBO: Robust Deep Reinforcement Learning via Multi-View Information Bottleneck
• Learning State Representations from Random Deep Action-conditional Predictions