Support Vector Regression Based on Adjustable Entropy Function Approach

Support vector machine is an elegant tool for solving pattern recognition and regression problems. This paper presents a new smooth approach to solve support vector regression. Based on statistical learning theory and optimization theory, a smooth unconstrained optimization model for support vector regression is built with adjustable entropy technique. Newton descent method is used to solve the model. The proposed approach can overcome the numerical overflow in the traditional entropy function approaches. Primary numerical results illustrate that our proposed approach improves the regression performance and the learning efficiency.