Large margin classification using the perceptron algorithm

摘要: We introduce and analyze a new algorithm for linear classification which combines Rosenblatt‘s perceptron with Helmbold Warmuth‘s leave-one-out method. Like Vapnik‘s maximal-margin classifier, our takes advantage of data that are linearly separable large margins. Compared to algorithm, however, ours is much simpler implement, more efficient in terms computation time. also show can be efficiently used very high dimensional spaces using kernel functions. performed some experiments variants it, classifying images handwritten digits. The performance close to, but not as good as, the classifiers on same problem, while saving significantly time programming effort.