Asymptotically Optimal Algorithms for Budgeted Multiple Play Bandits [article]

Alexander Luedtke, Emilie Kaufmann, Antoine Chambaz (MAP5 - UMR 8145)
2019 arXiv   pre-print
We study a generalization of the multi-armed bandit problem with multiple plays where there is a cost associated with pulling each arm and the agent has a budget at each time that dictates how much she can expect to spend. We derive an asymptotic regret lower bound for any uniformly efficient algorithm in our setting. We then study a variant of Thompson sampling for Bernoulli rewards and a variant of KL-UCB for both single-parameter exponential families and bounded, finitely supported rewards.
more » ... e show these algorithms are asymptotically optimal, both in rateand leading problem-dependent constants, including in the thick margin setting where multiple arms fall on the decision boundary.
arXiv:1606.09388v3 fatcat:uek4qx7oavhc7lkx4hydfifajy