A Modified Genetic Algorithm for Training the Standard Support Vector Machine

The standard problem (S P) for support vector machines (S VMs), proposed by Fletcher and Zanghirati, is a nonconvex nonlinear programming problem, and the general optimization methods in training the S P may obtain a local optimal solution. In order to avoid the local optimal solution, this paper utilizes a kind of modified genetic algorithm (MGA) to search the global optimal solution of the S P. At the same time, the MGA fully exploits the superiority and characteristic of the S P, which indicates that our method is effective and feasible. Some numerical experiments are also given, which illustrate that our genetic algorithm is effective.