A Note on Information-Directed Sampling and Thompson Sampling

TL;DR

This paper provides an intuitive explanation and derivations for three Bayesian multi-armed bandit algorithms: Information-directed sampling, Thompson Sampling, and Generalized Thompson Sampling, focusing on their regret bounds.

Contribution

It offers clear explanations and derivations for these algorithms, enhancing understanding of their theoretical properties.

Findings

Provides regret bounds for the algorithms

Clarifies the intuition behind each method

Includes derivations omitted in original works

Abstract

This note introduce three Bayesian style Multi-armed bandit algorithms: Information-directed sampling, Thompson Sampling and Generalized Thompson Sampling. The goal is to give an intuitive explanation for these three algorithms and their regret bounds, and provide some derivations that are omitted in the original papers.