Smooth twin bounded support vector machine with pinball loss

Abstract The twin support vector machine improves the classification performance of by solving two small quadratic programming problems. However, this method has following defects: (1) For and some its variants, constructed models use a hinge loss function, which is sensitive to noise unstable in resampling. (2) need be converted from original space dual space, their time complexity high. To further enhance machine, pinball function introduced into bounded problem not being differentiable at zero solved constructing smooth approximation function. Based on this, model with obtained. iteratively using Newton-Armijo method. A algorithm proposed, theoretically convergence iterative proven. In experiments, proposed validated UCI datasets artificial datasets. Furthermore, presented compared those other representative algorithms, thereby demonstrating effectiveness algorithm.