Many classifiers were designed, based on the assumption of normal distributions, kernels or potential functions, nearest neighbor rules,multi-layer perceptrons, and so on [15, 45, 62, 76].