A practical algorithm for multiplayer bandits when arm means vary among players
A Mehrabian, E Boursier… - International …, 2020 - proceedings.mlr.press
International Conference on Artificial Intelligence and Statistics, 2020•proceedings.mlr.press
We study a multiplayer stochastic multi-armed bandit problem in which players cannot
communicate, and if two or more players pull the same arm, a collision occurs and the
involved players receive zero reward. We consider the challenging heterogeneous setting,
in which different arms may have different means for different players, and propose a new
and efficient algorithm that combines the idea of leveraging forced collisions for implicit
communication and that of performing matching eliminations. We present a finite-time …
communicate, and if two or more players pull the same arm, a collision occurs and the
involved players receive zero reward. We consider the challenging heterogeneous setting,
in which different arms may have different means for different players, and propose a new
and efficient algorithm that combines the idea of leveraging forced collisions for implicit
communication and that of performing matching eliminations. We present a finite-time …
Abstract
We study a multiplayer stochastic multi-armed bandit problem in which players cannot communicate, and if two or more players pull the same arm, a collision occurs and the involved players receive zero reward. We consider the challenging heterogeneous setting, in which different arms may have different means for different players, and propose a new and efficient algorithm that combines the idea of leveraging forced collisions for implicit communication and that of performing matching eliminations. We present a finite-time analysis of our algorithm, giving the first sublinear minimax regret bound for this problem, and prove that if the optimal assignment of players to arms is unique, our algorithm attains the optimal O (ln (T)) regret, solving an open question raised at NeurIPS 2018 by Bistritz and Leshem (2018).
proceedings.mlr.press
顯示最佳搜尋結果。 查看所有結果