PAC-Bayesian Learning of Linear Classifiers

We present a general PAC-Bayes theorem from which all known PAC-Bayes bounds are simply obtained as particular cases. We also propose different learning algorithms for finding linear classifiers that