Multi-armed Bandit Problems with History


Pannagadatta Shivaswamy, Thorsten Joachims ;
Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, PMLR 22:1046-1054, 2012.


In this paper we consider the stochastic multi-armed bandit problem. However, unlike in the conventional version of this problem, we do not assume that the algorithm starts from scratch. Many applications offer observations of (some of) the arms even before the algorithm starts. We propose three novel multi-armed bandit algorithms that can exploit this data. An upper bound on the regret is derived in each case. The results show that a logarithmic amount of historic data can reduce regret from logarithmic to constant. The effectiveness of the proposed algorithms are demonstrated on a large-scale malicious URL detection problem.

Related Material