Tsetlin Machine for Solving Contextual Bandit Problems

TL;DR

This paper presents an interpretable Tsetlin Machine-based algorithm for contextual bandit problems, demonstrating superior performance on multiple datasets and providing insights into decision interpretability through propositional logic.

Contribution

It introduces a novel Tsetlin Machine approach for contextual bandits, integrating Thompson sampling and emphasizing interpretability and computational simplicity.

Findings

Outperforms other learners on 8 of 9 datasets

Provides interpretable propositional logic explanations

Simplifies computation with bit manipulation

Abstract

This paper introduces an interpretable contextual bandit algorithm using Tsetlin Machines, which solves complex pattern recognition tasks using propositional logic. The proposed bandit learning algorithm relies on straightforward bit manipulation, thus simplifying computation and interpretation. We then present a mechanism for performing Thompson sampling with Tsetlin Machine, given its non-parametric nature. Our empirical analysis shows that Tsetlin Machine as a base contextual bandit learner outperforms other popular base learners on eight out of nine datasets. We further analyze the interpretability of our learner, investigating how arms are selected based on propositional expressions that model the context.