Taming the Monster: A Fast and Simple Algorithm for Contextual Bandits
This provides a practical solution for contextual bandit learning, applicable to general policy classes, though it appears incremental in improving efficiency.
The paper tackles the contextual bandit problem by introducing a fast and simple algorithm that achieves statistically optimal regret with only $ ilde{O}(\sqrt{KT/\log N})$ oracle calls, and experiments show it outperforms baselines in computational and prediction performance.
We present a new algorithm for the contextual bandit learning problem, where the learner repeatedly takes one of $K$ actions in response to the observed context, and observes the reward only for that chosen action. Our method assumes access to an oracle for solving fully supervised cost-sensitive classification problems and achieves the statistically optimal regret guarantee with only $\tilde{O}(\sqrt{KT/\log N})$ oracle calls across all $T$ rounds, where $N$ is the number of policies in the policy class we compete against. By doing so, we obtain the most practical contextual bandit learning algorithm amongst approaches that work for general policy classes. We further conduct a proof-of-concept experiment which demonstrates the excellent computational and prediction performance of (an online variant of) our algorithm relative to several baselines.