Diverse Projection Ensembles for Distributional Reinforcement Learning
This work addresses the challenge of reliable uncertainty estimation in distributional RL, which is crucial for improving exploration in reinforcement learning tasks, though it is incremental as it builds on existing projection methods.
The paper tackles the problem of strong inductive bias in distributional reinforcement learning by proposing diverse projection ensembles, which combine different projections and representations to improve uncertainty estimation and exploration. The algorithm achieved significant performance improvements on the behavior suite benchmark and VizDoom, with the most pronounced gains in directed exploration problems.
In contrast to classical reinforcement learning (RL), distributional RL algorithms aim to learn the distribution of returns rather than their expected value. Since the nature of the return distribution is generally unknown a priori or arbitrarily complex, a common approach finds approximations within a set of representable, parametric distributions. Typically, this involves a projection of the unconstrained distribution onto the set of simplified distributions. We argue that this projection step entails a strong inductive bias when coupled with neural networks and gradient descent, thereby profoundly impacting the generalization behavior of learned models. In order to facilitate reliable uncertainty estimation through diversity, we study the combination of several different projections and representations in a distributional ensemble. We establish theoretical properties of such projection ensembles and derive an algorithm that uses ensemble disagreement, measured by the average 1-Wasserstein distance, as a bonus for deep exploration. We evaluate our algorithm on the behavior suite benchmark and VizDoom and find that diverse projection ensembles lead to significant performance improvements over existing methods on a variety of tasks with the most pronounced gains in directed exploration problems.