Never Give Up: Learning Directed Exploration Strategies

Abstract

We propose a reinforcement learning agent to solve hard exploration games by learning a range of directed exploratory policies. We construct an episodic memory-based intrinsic reward using k-nearest neighbors over the agent's recent experience to train the directed exploratory policies, thereby encouraging the agent to repeatedly revisit all states in its environment. A self-supervised inverse dynamics model is used to train the embeddings of the nearest neighbour lookup, biasing the novelty signal towards what the agent can control. We employ the framework of Universal Value Function Approximators (UVFA) to simultaneously learn many directed exploration policies with the same neural network, with different trade-offs between exploration and exploitation. By using the same neural network for different degrees of exploration/exploitation, transfer is demonstrated from predominantly exploratory policies yielding effective exploitative policies. The proposed method can be incorporated to run with modern distributed RL agents that collect large amounts of experience from many actors running in parallel on separate environment instances. Our method doubles the performance of the base agent in all hard exploration in the Atari-57 suite while maintaining a very high score across the remaining games, obtaining a median human normalised score of 1344.0%. Notably, the proposed method is the first algorithm to achieve non-zero rewards (with a mean score of 8,400) in the game of Pitfall! without using demonstrations or hand-crafted features.

Published as a conference paper in ICLR 2020

References in corpus (13)

• Dueling Network Architectures for Deep Reinforcement Learning
• IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
• Distributed Prioritized Experience Replay
• Large-Scale Study of Curiosity-Driven Learning
• Distributed Distributional Deterministic Policy Gradients
• Incentivizing Exploration In Reinforcement Learning With Deep Predictive Models
• Exploration by Random Network Distillation
• Go-Explore: a New Approach for Hard-Exploration Problems
• Episodic Curiosity through Reachability
• Model-Free Episodic Control
• Observe and Look Further: Achieving Consistent Performance on Atari
• Deep Curiosity Search: Intra-Life Exploration Can Improve Performance on Challenging Deep Reinforcement Learning Problems
• MULEX: Disentangling Exploitation from Exploration in Deep RL

Cited by in corpus (32)

• Agent57: Outperforming the Atari Human Benchmark
• First return, then explore
• An information-theoretic perspective on intrinsic motivation in reinforcement learning: a survey
• Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learning
• Grounded Language Learning Fast and Slow
• Human-in-the-Loop Deep Reinforcement Learning with Application to Autonomous Driving
• Variational Dynamic for Self-Supervised Exploration in Deep Reinforcement Learning
• State Entropy Maximization with Random Encoders for Efficient Exploration
• Dynamic Bottleneck for Robust Self-Supervised Exploration
• Beyond Fine-Tuning: Transferring Behavior in Reinforcement Learning
• Behavior From the Void: Unsupervised Active Pre-Training
• MADE: Exploration via Maximizing Deviation from Explored Regions
• The Advantage Regret-Matching Actor-Critic
• Self-Imitation Advantage Learning
• Temporal Difference Uncertainties as a Signal for Exploration
• Principled Exploration via Optimistic Bootstrapping and Backward Induction
• Novelty Search in Representational Space for Sample Efficient Exploration
• Return-based Scaling: Yet Another Normalisation Trick for Deep RL
• Rank the Episodes: A Simple Approach for Exploration in Procedurally-Generated Environments
• Disentangling causal effects for hierarchical reinforcement learning
• Uncertainty-Aware Model-Based Reinforcement Learning with Application to Autonomous Driving
• Group Equivariant Deep Reinforcement Learning
• Temporally-Extended ε-Greedy Exploration
• Don't Do What Doesn't Matter: Intrinsic Motivation with Action Usefulness
• Adapting to Reward Progressivity via Spectral Reinforcement Learning
• Policy Manifold Search for Improving Diversity-based Neuroevolution
• Did I do that? Blame as a means to identify controlled effects in reinforcement learning
• When should agents explore?
• Inducing Functions through Reinforcement Learning without Task Specification
• More Efficient Exploration with Symbolic Priors on Action Sequence Equivalences
• Eden: A Unified Environment Framework for Booming Reinforcement Learning Algorithms
• Decoupled Exploration and Exploitation Policies for Sample-Efficient Reinforcement Learning