Reviews: Quantifying Learning Guarantees for Convex but Inconsistent Surrogates

Summary The paper provides a new lower bound in consistency analysis of convex surrogate loss functions. Section 1 provides an introduction which discusses previous work and contributions. The main previous work is Osokin et al, which is frequently referenced in the text. The main contribution is a generalization of the results of Osokin et al to inconsistent surrogates, and a new tighter lower bound. An additional contribution is a study the behavior of the proposed bound for two prediction problems.