Empirical Likelihood for Contextual Bandits
This work addresses the challenge of reliable off-policy evaluation and optimization in contextual bandits, which is incremental but offers practical improvements for reinforcement learning applications.
The authors tackled the problem of estimating policy value and confidence intervals from off-policy data in contextual bandits, proposing an empirical likelihood-based estimator and confidence interval that improve over previous methods in finite sample regimes, and an optimization algorithm using these intervals that outperforms a strong baseline.
We propose an estimator and confidence interval for computing the value of a policy from off-policy data in the contextual bandit setting. To this end we apply empirical likelihood techniques to formulate our estimator and confidence interval as simple convex optimization problems. Using the lower bound of our confidence interval, we then propose an off-policy policy optimization algorithm that searches for policies with large reward lower bound. We empirically find that both our estimator and confidence interval improve over previous proposals in finite sample regimes. Finally, the policy optimization algorithm we propose outperforms a strong baseline system for learning from off-policy data.