Kernel-parameter Selection Problem in Support Vector Machine

Kernel function and kernel-parameter play an important role in Support Vector Machine.It has a great influence on the model application.Minimizing the LOO upper bound is one of efficient method to select kernel-parameter.Usually,the steepest descent algorithm is used to find the minimum of LOO upper-bound.However,it often gets local optimal solution.To overcome this shortcoming,based on hybrid genetic algorithm a new kernel-parameter selection method is proposed.Furthermore,some experiments have been made and the results show that the new method is very efficient.