Data di Pubblicazione:
2024
Citazione:
Sum-max Submodular Bandits / S. Pasteris, A. Rumi, F. Vitale, N. Cesa Bianchi (PROCEEDINGS OF MACHINE LEARNING RESEARCH). - In: Proceedings of The 27th International Conference on Artificial Intelligence and Statistics / [a cura di] S. Dasgupta, S. Mandt, Y. Li. - [s.l] : ML Research Press, 2024. - pp. 2323-2331 (( Intervento presentato al 27. convegno International Conference on Artificial Intelligence and Statistics tenutosi a Valencia nel 2024.
Abstract:
Many online decision-making problems correspond to maximizing a sequence of submodular functions. In this work, we introduce sum-max functions, a subclass of monotone submodular functions capturing several interesting problems, including best-of-$K$-bandits, combinatorial bandits, and the bandit versions on $M$-medians and hitting sets. We show that all functions in this class satisfy a key property that we call pseudo-concavity. This allows us to prove $\big(1 - \frac{1}{e}\big)$-regret bounds for bandit feedback in the nonstochastic setting of the order of $\sqrt{MKT}$ (ignoring log factors), where $T$ is the time horizon and $M$ is a cardinality constraint. This bound, attained by a simple and efficient algorithm, significantly improves on the $\widetilde{\mathcal{O}}\big(T^{2/3}\big)$ regret bound for online monotone submodular maximization with bandit feedback. We also extend our results to a bandit version of the facility location problem.
Tipologia IRIS:
03 - Contributo in volume
Elenco autori:
S. Pasteris, A. Rumi, F. Vitale, N. Cesa Bianchi
Link alla scheda completa:
Link al Full Text:
Titolo del libro:
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics