Uncoupled mixture probability density estimation based on an improved support vector machine model

Abstract

Support vector machine(SVM) is a new approach for probability density estimation problems. But there are some shortcomings in the SVM based method, for example, the method can only optimize the model directly, and the slack factors must belong to the optimized range of solutions. On this basis, an improved SVM model named single slack factor SVM probability density estimation model is proposed in the paper. In this model, the scale of object function is reduced, so the computation efficient is greatly enhanced. The experiment results on uncoupled mixture probability density estimation show the effectiveness and feasibility of the model.