A Bayesian Sampling Approach to Exploration in Reinforcement Learning
This addresses the exploration challenge in reinforcement learning for AI agents, offering a modular and flexible method that is competitive with state-of-the-art approaches.
The paper tackles the exploration problem in reinforcement learning by introducing BOSS, a Bayesian sampling approach that selects actions optimistically from sampled models, achieving near-optimal reward with high probability and low sample complexity relative to posterior convergence.
We present a modular approach to reinforcement learning that uses a Bayesian representation of the uncertainty over models. The approach, BOSS (Best of Sampled Set), drives exploration by sampling multiple models from the posterior and selecting actions optimistically. It extends previous work by providing a rule for deciding when to resample and how to combine the models. We show that our algorithm achieves nearoptimal reward with high probability with a sample complexity that is low relative to the speed at which the posterior distribution converges during learning. We demonstrate that BOSS performs quite favorably compared to state-of-the-art reinforcement-learning approaches and illustrate its flexibility by pairing it with a non-parametric model that generalizes across states.