Interpolation based kernel function´s constructing

The kernel function is important for support vector machines (SVMs) in classification and regression. However, how to select a kernel function for the given data is still an open problem. Many papers are limited to consult the properties of some standard kernel functions. Since the effect of kernel mapping has not been understood very clearly, the result may not be as good as SVM should be in some cases. We present a method with interpolation and the subspace method to construct a kernel function according to the given data. The experiments showed that the kernel function constructed using our method has predominance of generalization in the case that the training data is too small.