Research for face recognition base on mixed kernel function
Shuxian Zhu, Renjie Zhang · 2008
Support Vector Machines were developed in recent years, which have large advantage over the traditional neural network on small sample set for classification. In all research fields of these learning machines, the selection of kernel function is the most important problem, which has a closed relationship with the performance of classification. But the research work in this field is not enough. In this paper we evaluate the performance of usual kernel functions for SVM theoretically, through observing and computing the kernel matrix. Base on this, we used the selected kernel functions to get a new mixed kernel function. Experiential data proved that the performance of SVM was improved by the mixed kernel function. If we select the weighted values properly, the correct rate even is 100%. This will not only gives us a method to get a new learning machine, but also give a reference for selecting kernel function.