Tactical Optimism and Pessimism for Deep Reinforcement Learning
This work provides a method to improve the performance of deep reinforcement learning algorithms for continuous control by adaptively balancing exploration and stability.
This paper addresses the trade-off between optimism for exploration and pessimism for stability in deep off-policy actor-critic algorithms. The authors introduce Tactical Optimistic and Pessimistic (TOP) estimation, a framework that dynamically switches between optimistic and pessimistic value learning, achieving new state-of-the-art performance in challenging pixel-based continuous control environments.
In recent years, deep off-policy actor-critic algorithms have become a dominant approach to reinforcement learning for continuous control. One of the primary drivers of this improved performance is the use of pessimistic value updates to address function approximation errors, which previously led to disappointing performance. However, a direct consequence of pessimism is reduced exploration, running counter to theoretical support for the efficacy of optimism in the face of uncertainty. So which approach is best? In this work, we show that the most effective degree of optimism can vary both across tasks and over the course of learning. Inspired by this insight, we introduce a novel deep actor-critic framework, Tactical Optimistic and Pessimistic (TOP) estimation, which switches between optimistic and pessimistic value learning online. This is achieved by formulating the selection as a multi-arm bandit problem. We show in a series of continuous control tasks that TOP outperforms existing methods which rely on a fixed degree of optimism, setting a new state of the art in challenging pixel-based environments. Since our changes are simple to implement, we believe these insights can easily be incorporated into a multitude of off-policy algorithms.