Dynamic Bottleneck for Robust Self-Supervised Exploration

Abstract

Exploration methods based on pseudo-count of transitions or curiosity of dynamics have achieved promising results in solving reinforcement learning with sparse rewards. However, such methods are usually sensitive to environmental dynamics-irrelevant information, e.g., white-noise. To handle such dynamics-irrelevant information, we propose a Dynamic Bottleneck (DB) model, which attains a dynamics-relevant representation based on the information-bottleneck principle. Based on the DB model, we further propose DB-bonus, which encourages the agent to explore state-action pairs with high information gain. We establish theoretical connections between the proposed DB-bonus, the upper confidence bound (UCB) for linear case, and the visiting count for tabular case. We evaluate the proposed method on Atari suits with dynamics-irrelevant noises. Our experiments show that exploration with DB bonus outperforms several state-of-the-art exploration methods in noisy environments.

NeurIPS 2021

References in corpus (5)

• Bootstrap your own latent: A new approach to self-supervised Learning
• Never Give Up: Learning Directed Exploration Strategies
• On Bonus-Based Exploration Methods in the Arcade Learning Environment
• Optimistic Exploration even with a Pessimistic Initialisation
• APS: Active Pretraining with Successor Features