Policy Gradient Optimization of Thompson Sampling Policies
This work addresses a known limitation in Thompson sampling for sequential decision-making, offering an incremental enhancement to a widely used algorithm in reinforcement learning and bandit problems.
The paper tackles the problem of improving Thompson sampling in Bayesian bandit settings by using policy gradient methods to optimize over a class of generalized Thompson sampling policies, resulting in meaningful performance improvements even in long-horizon problems.
We study the use of policy gradient algorithms to optimize over a class of generalized Thompson sampling policies. Our central insight is to view the posterior parameter sampled by Thompson sampling as a kind of pseudo-action. Policy gradient methods can then be tractably applied to search over a class of sampling policies, which determine a probability distribution over pseudo-actions (i.e., sampled parameters) as a function of observed data. We also propose and compare policy gradient estimators that are specialized to Bayesian bandit problems. Numerical experiments demonstrate that direct policy search on top of Thompson sampling automatically corrects for some of the algorithm's known shortcomings and offers meaningful improvements even in long horizon problems where standard Thompson sampling is extremely effective.