arXiv Analytics

Sign in

arXiv:2206.11254 [cs.LG]AbstractReferencesReviewsResources

Langevin Monte Carlo for Contextual Bandits

Pan Xu, Hongkai Zheng, Eric Mazumdar, Kamyar Azizzadenesheli, Anima Anandkumar

Published 2022-06-22Version 1

We study the efficiency of Thompson sampling for contextual bandits. Existing Thompson sampling-based algorithms need to construct a Laplace approximation (i.e., a Gaussian distribution) of the posterior distribution, which is inefficient to sample in high dimensional applications for general covariance matrices. Moreover, the Gaussian approximation may not be a good surrogate for the posterior distribution for general reward generating functions. We propose an efficient posterior sampling algorithm, viz., Langevin Monte Carlo Thompson Sampling (LMC-TS), that uses Markov Chain Monte Carlo (MCMC) methods to directly sample from the posterior distribution in contextual bandits. Our method is computationally efficient since it only needs to perform noisy gradient descent updates without constructing the Laplace approximation of the posterior distribution. We prove that the proposed algorithm achieves the same sublinear regret bound as the best Thompson sampling algorithms for a special case of contextual bandits, viz., linear contextual bandits. We conduct experiments on both synthetic data and real-world datasets on different contextual bandit models, which demonstrates that directly sampling from the posterior is both computationally efficient and competitive in performance.

Comments: 21 pages, 3 figures, 2 tables. To appear in the proceedings of the 39th International Conference on Machine Learning (ICML2022)
Categories: cs.LG, stat.ML
Related articles: Most relevant | Search more
arXiv:2006.05094 [cs.LG] (Published 2020-06-09)
Differentiable Meta-Learning in Contextual Bandits
arXiv:2001.07853 [cs.LG] (Published 2020-01-22)
Incentivising Exploration and Recommendations for Contextual Bandits with Payments
arXiv:1808.05760 [cs.LG] (Published 2018-08-17)
Data Poisoning Attacks in Contextual Bandits