Sun et al. study subset ranking using binary classifiers. They establish a reduction framework from subset ranking to binary classification, and establish new tight regret bounds extending and improving previous results. The paper shows that there is an underlying connection between generic subset ranking and binary classification. Specifically, an improvement in binary classification accuracy can enhance the ranking performance.

The paper focuses solely on the technical aspects of the studied problem and assumes that the reader has sufficient knowledge of the problem and its applications. The studied problem itself is never motivated or put into context. As such, this paper is primarily targeted at, and accessible to, a narrow audience of readers who are working on closely related problems. Other readers may unfortunately become quickly frustrated by the paper’s lack of exhibition.