Export Reviews, Discussions, Author Feedback and Meta-Reviews

First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. In this problem a subset S_t of available actions of a decision set S\subseteq {0,1}^d is chosen stochastically. Then the algorithm chooses an action v\in S_t and incurs a loss of v^T*l_t where l_t is a loss vector chosen by an oblivious adversary. The paper studies the problem in three settings full information, semi-bandit and a new setting which they term restricted. Results: Previously results were known in the full information and semi-bandit setting with sublinear regret bounds.