Abstract:

Using a class of new smoothing functions, the problem of smoothing support vector machines was studied. A new 3rdorder polynomial smooth support vector machine (3SSVM) model was proposed, and its 3rdorder smoothness and global convergence was proved. Numerical experiments were carried out to evaluate the 3SSVM by using NewtonArmijo algorithm. The results show that the 3SSVM model is better than the PSSVM model and the SSVM model, providing better theoretical support for applications of smooth support vector machines.

Key words: classification;support vector machine;polynomial;smoothing function;convergence