Online Learning Algorithm of Direct Support Vector Machine for Regression Based on Matrix Operation

摘要

With the development of support vector machine, its' application of online learning was paid close attention to more and more. In this paper ,we deduced an online learning algorithm for direct support vector machine (DSVM). By adding the bias square of hyperplane to the objective function, we introduced the mathematical mode of DSVM for regression which was extremely fast and simple, then we designed incremental and online learning algorithms by using the property of block matrix and kernel function matrix. Experimental results of simulation by using Mackey-Glass chaotic time series indicate the feasibility of the learning algorithm.