A PAC-Bayesian Margin Bound for Linear Classifiers: Why SVMs work

Open in new window