An effective algorithm for inverse problem of SVM based on MM algorithm

This paper investigates an effective algorithm for inverse problem of support vector machines. The inverse problem is how to split a given dataset into two clusters such that the margin between the two clusters attains maximum. However the training time for inverse problem of SVM is incredible. Clustering is a feasible way to simplify the process of it, but it is difficult to estimate the number of the clusters. In this paper, we design a margin-merging cluster algorithm to solve this problem. We compare our approach with the k-means solution in terms of accuracy loss and training time. Simulations show that the proposed algorithm can solve it efficiently.