The paper presents a supervised facility location based approach to subset selection, i.e., choosing a set of representative points from a new dataset. The paper considers a sparse convex relaxation of the problem and characterizes conditions for getting integral solutions. An alternating algorithm utilizing the integral solutions is presented for learning the subset mapping. Extensive experimental results are presented to illustrate the effectiveness of the proposed approach. The reviewers agree that the paper makes a novel contribution to an important problem and the paper is well written.