Beyond Optimism: Exploration With Partially Observable Rewards
This addresses a key challenge in RL for agents operating in environments with sparse or unobservable rewards, representing an incremental advance over optimism-based approaches.
The paper tackles the problem of exploration in reinforcement learning when rewards are partially observable, presenting a novel exploration strategy that guarantees convergence to an optimal policy and outperforms existing methods on new benchmark environments.
Exploration in reinforcement learning (RL) remains an open challenge. RL algorithms rely on observing rewards to train the agent, and if informative rewards are sparse the agent learns slowly or may not learn at all. To improve exploration and reward discovery, popular algorithms rely on optimism. But what if sometimes rewards are unobservable, e.g., situations of partial monitoring in bandits and the recent formalism of monitored Markov decision process? In this case, optimism can lead to suboptimal behavior that does not explore further to collapse uncertainty. With this paper, we present a novel exploration strategy that overcomes the limitations of existing methods and guarantees convergence to an optimal policy even when rewards are not always observable. We further propose a collection of tabular environments for benchmarking exploration in RL (with and without unobservable rewards) and show that our method outperforms existing ones.