A Non-linear Classifier Based on the Contraction of the Closed Convex Hull

摘要

In this paper, the bisecting-nearest-point method is extended and transformed to a non-linear classifier method utilizing the kernel theory. As for the nonlinear inseparability of classification, the contraction of a closed convex hull algorithm in feature space is put forward, which can turn inseparability into separability by properly contracting the specimen in feature space. The algorithm proposed in this paper possesses not only simpler and more intuitionistic geometric meaning but also the same effect as SVM in classifying capability, and can also effectively decrease the computing complexity of classifying hyperplane.