Online Improper Learning with an Approximation Oracle

Elad Hazan, Wei Hu, Yuanzhi Li, Zhiyuan Li

Neural Information Processing Systems 

We study the following question: given an efficient approximation algorithm for an optimization problem, can we learn efficiently in the same setting? We give a formal affirmative answer to this question in the form of a reduction from online learning to offline approximate optimization using an efficient algorithm that guarantees near optimal regret. The algorithm is efficient in terms of the number of oracle calls to a given approximation oracle - it makes only logarithmically many such calls per iteration.