The Research on the Multiple Kernel Learning-based Face Recognition inPattern Matching

Abstract

The paper analyses the multiple kernel learning-based face recognition in pattern matching area. Based on the analysis of the basic theory of multiple kernel SVM, this thesis focuses on the multiple kernel SVM algorithm based on semi-infinite linear program (SILP), including SILP based on column generation (CG) and SILP based on chunking algo- rithm (CA). The two SILP improved algorithms are applied to several classification problems, including UCI binary clas- sification problem datasets and multi-classification problem datasets. Furthermore, the two SILP improved algorithms are applied to the actual problems of face recognition. The experiment data shows that with the multiple kernel learning-based method, the performance of face recognition can be obviously improved.