Topology potential-based parameter selecting for support vector machine

We present an algorithm for selecting support vector machine’s meta- parameter value which is based on ideas from topology potential of data field. By the optimal spatial distribution of topological potential corresponding to minimum entropy potential, it searches so smart that the optimal parameters can be found effectively and efficiently. The experimental results show that it can get almost the same effectiveness with the exhaustive grid search under an order of magnitude lower computational cost. It also can be used to automatically identify kernels and other parameter selection problem.