Capped Asymmetric Elastic Net Support Vector Machine for Robust Binary Classification

Recently, there are lots of literature on improving the robustness SVM by constructing nonconvex functions, but they seldom theoretically study robust property constructed functions. In this paper, based our recent work, we present a novel capped asymmetric elastic net (CaEN) loss and equip it with as CaENSVM. We derive influence function estimators CaENSVM to explain proposed method. Our results can be easily extended other similar further show that is bounded, so explained. Other theoretical analysis demonstrates satisfies Bayes rule corresponding generalization error bound Rademacher complexity guarantees its good capability. Since CaEN concave, implement an efficient DC procedure stochastic gradient descent algorithm (Pegasos) solve optimization problem. A host experiments conducted verify effectiveness model.