The hinge loss is one-Lipschitz. Centering it at zero and applying the contraction lemma gives
For an empirical risk minimizer and a population minimizer, the standard expected ERM inequality with the Rademacher convention used in part (a) is
Part (c) consequently yields
Thus for this normalization of Rademacher complexity.

Articles by others on the same topic (0)

There are currently no matching articles.