Least Squares Support Matrix Machine Based On Bi-Level Programming And Its Fast Algorithm

At first,this paper will discuss the advantages of many kinds of least square support vector machines(LSSVM),then least square support matrix machines based on bi-level programming(BP-LSSMM)and quickly least square support matrix machines are based on bi-level programming(Q-BP-LSSMM)will be proposed.The content is divided into four parts.In the first chapter,we will give a brief review of kernel function?SVM?SVR?TSVM and SMM.In the second chapter,eight different kinds SVM and TSVM are based on least square technique will be introduced,comparing their different performances of many kinds LSSVM model Learning algorithm.In the third chapter,we will give the model of BP-LSSMM and put award learning algorithm.What is more,through experiment proves the efficient of this model.BP-LSSMM has the advantages of two aspects,one is for input,it is the matrix sample.We don't need to quantify it and we can learn it directly,avoid dimension disaster.The other side uses the advantage of least squares techniques can improve the speed under the premise of guarantee the accuracy.In the fourth chapter utilizes the idea of bi-level programming and swaps the upper problem and the lower problem of double problem based on BP-LSSMM,it is a timesaver learning algorithm and verified the validity of the proposed algorithm through the experiment.