Reviews: Doubly-Robust Lasso Bandit

Neural Information Processing Systems 

The exposition is quite unclear, and the paper seems hastily written. The main contributions of the paper is not outlined clearly, and not compared rigorously with existing results (see below). I believe there are enough theoretical contributions in the paper, but do not recommend publication as is. Ideally, I would recommend a "revise and resubmit". Even when presented in the best light, the main regret bound gives a significant deterioration in terms of its dependence on T, which is concerning.