A New Class of epsilon-Piecewise Smooth Support Vector Regressions

作者：Wu Qing^*

来源：Journal of Information Science and Engineering, 2015, 31(5): 1813-1828.

• support vector regression
• smooth technique
• piecewise function
• Newton-Armijo method

epsilon-support vector regression (epsilon-SVR) can be converted into an unconstrained convex and non-smooth quadratic programming problem. It is not solved by the traditional algorithm. In order to solve this non-smooth problem, a class of piecewise smooth functions is introduced to approximate the epsilon-insensitive loss function of epsilon-SVR, which generates a epsilon-piecewise smooth support vector regression (epsilon-dPWSSVR) model. The fast Newton-Armijo algorithm is used to solve the epsilon-dPWSSVR. The piecewise functions can get higher and higher approximation accuracy as required with increase of parameter d. The reduced kernel technique is applied to avoid the computational difficulties in nonlinear epsilon-dPWSSVR for massive datasets. Experimental results show that the proposed epsilon-dPWSSVR has the better regression performance and the learning efficiency than other competitive baselines.

• 出版日期2015-9
• 单位西安邮电大学