Combinatorial Bandits Revisited

Author name not available (Why is that?)

Publication date: 11 February 2015

Abstract: This paper investigates stochastic and adversarial combinatorial multi-armed bandit problems. In the stochastic setting under semi-bandit feedback, we derive a problem-specific regret lower bound, and discuss its scaling with the dimension of the decision space. We propose ESCB, an algorithm that efficiently exploits the structure of the problem and provide a finite-time analysis of its regret. ESCB has better performance guarantees than existing algorithms, and significantly outperforms these algorithms in practice. In the adversarial setting under bandit feedback, we propose extsc{CombEXP}, an algorithm with the same regret scaling as state-of-the-art algorithms, but with lower computational complexity for some combinatorial problems.

Has companion code repository: https://github.com/gitting-guud/GML_Project

This page was built for publication: Combinatorial Bandits Revisited

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6259005)