Imitation-Regularized Offline Learning

Yifei Ma, Yu-Xiang Wang, Balakrishnan +1

We study the problem of offline learning in automated decision systems under the contextual bandits model. We are given logged historical data consisting of contexts, (randomized)…