Thompson Sampling with Approximate Inference

Abstract
We study the effects of approximate inference on the performance of Thompson sampling in the -armed bandit problems. Thompson sampling is a successful algorithm for online decision-making but requires posterior inference, which often must be approximated in practice. We show that even small constant inference error (in -divergence) can lead to poor performance (linear regret) due to under-exploration (for ) or over-exploration (for ) by the approximation. While for this is unavoidable, for the regret can be improved by adding a small amount of forced exploration even when the inference error is a large constant.
View on arXivComments on this paper