[Home]


STPMSVM

A Matlab code for smooth twin parametric-margin support vector machine.(You could Right-Click [Code] , and Save, then you can download the whole matlab code.)


Reference

Zhen Wang, Yuan-Hai Shao, Tie-Ru Wu. A GA-based model selection for smooth twin parametric-margin support vector machine[J]. Pattern Recognition, 46(8) (2013) 2267-2277.


Main Function

This is the linear version STPMSVM, the nonlingear one can be down load from the [Code].

function testY = Stpmsvm(testX,DataTrain,FunPara) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % STPMSVM: linear smooth twin parametric-margin support vector machine % % testY = Stpmsvm(testX,DataTrain,FunPara) % % Input: % testX - test Data matrix. Each row vector of fea is a data point. % % DataTrain - Struct value in Matlab(Training data). % DataTrain.A: Positive input of Data matrix. % DataTrain.B: Negative input of Data matrix. % % FunPara - Struct value in Matlab. The fields in options that can be set: % c1: [0,inf] Paramter to tune the weight. % c2: [0,inf] Paramter to tune the weight. % v1: [0,inf] Paramter to tune the weight. % v2: [0,inf] Paramter to tune the weight. % % Output: % test_Y - Predict value of the TestX. % % Examples: % DataTrain.A = rand(50,10); % DataTrain.B = rand(60,10); % testX=rand(20,10); % FunPara.c1=0.1; % FunPara.c2=0.1; % FunPara.v1=0.1; % FunPara.v2=0.1; % testY = Stpmsvm(testX,DataTrain,FunPara); % % Reference: % Zhen Wang, Yuan-Hai Shao, Tie-Ru Wu.A GA-based model selection for % smooth twin parametric-margin support vector machine % Pattern Recognition, 46(8) (2013) 2267C2277 % % Version 1.0 --Apr/2013 % % Written by Zhen-Wang (wangzhen1882@126.com) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Initailization %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %tic; inputA = DataTrain.A; inputB = DataTrain.B; c1 = FunPara.c1; c2 = FunPara.c2; v1 = FunPara.v1; v2 = FunPara.v2; A=[inputA,ones(size(inputA,1),1)]; B=[inputB,ones(size(inputB,1),1)]; w1=-ones(size(A,2),1); %w0 w2=-ones(size(B,2),1); %w0 tol=0.001; %eps Max=1000; % iteration max count %training [w1,w2]=train(A,B,c1,c2,v1,v2,w1,w2,tol,Max); %toc %testing testY=test(w1,w2,testX); end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Training function %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function [w1,w2]=train(A,B,c1,c2,v1,v2,w1,w2,tol,max_ite) % training process [n1,m1]=size(A); [n2,m2]=size(B); e1=ones(n1,1); %Aeq e2=ones(n2,1); % solve first optimal problem flag=1; tol0=tol; ite=0; while flag>tol0 && itetol0 % compute Hessian index=find(plus>0); Q=eye(m1)+c1/n1*A(index,:)'*A(index,:); % Newton stepsize z=Q\(-grad); w1=w1+z; else flag=tol0; end end % solve second optimal problem flag=1; tol0=tol; ite=0; while flag>tol0 && itetol0 % compute Hessian index=find(plus>0); Q=eye(m2)+c2/n2*B(index,:)'*B(index,:); % Newton stepsize z=Q\(-grad); w2=w2+z; else flag=tol0; end end end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Predict function %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% function Y=test(w1,w2,X) % testing process % w1,w2 all have been considered the bias 'b'! % X not considered the bias 'e' [m,n]=size(w1); norm1=norm(w1(1:(m-1),:),2); norm2=norm(w2(1:(m-1),:),2); Y=sign(X*(w1(1:(m-1),1)/norm1+w2(1:(m-1),1)/norm2)+(w1(m,1)/norm1+w2(m,1)/norm2)); end
Contacts


Any question or advice please email to wangzhen1882@126.com and shaoyuanhai21@163.com