A New Data-dependent Kernel Intelligent Optimization Method

In order to deal with the kernel optimization,a new intelligent data-dependent optimizing kernel method is proposed.In this scheme,a new kernel cluster balanced K-means is firstly presented,which can effectively overcome the singularity of matrix for Mahalanobis distance in local Fisher criterion(KO-LKFC).Then,the total objective function based on local Fisher criterion is given.Finally,genetic algorithm is used to find the global optimization and the intelligence is improved.Experimental results based on UCI data and the MSTAR SAR data show that it is effective and feasible.