PCC

A Matlab code for proximal classifier with consistency. (You could Right-Click [Code] , and Save, then you can download the whole matlab code.)

Reference

Y.-H. Shao, N.-Y. Deng, W.-J. Chen. A proximal classifier with consistency, Knowledge-Based Systems,2013, 49:171-178

Main Function

function Predict_Y = PCC(TestX,DataTrain,FunPara) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % PCC: A proximal classifier with consistency % % Predict_Y = PCC(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: % c: [0,inf] Paramter to tune the weight. % mu: [0,inf] Paramter to tune the weight. % kerfPara:Kernel parameters. See kernelfun.m. % % Output: % Predict_Y - Predict value of the TestX. % % Examples: % DataTrain.A = rand(50,10); % DataTrain.B = rand(60,10); % TestX=rand(20,10); % FunPara.c=0.1; % FunPara.mu=0.1; % FunPara.kerfPara.type = 'lin'; % Predict_Y =PCC(TestX,DataTrain,FunPara); % % Reference: % Yuan-Hai Shao, Nai-Yang Deng and Wei-Jie Chen, "A proximal classifier % with consistency" Submitted 2013 % % Version 1.0 --Apr/2013 % % Written by Wei-Jie Chen (wjcper2008@126.com) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Initailization %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %tic; Xpos = DataTrain.A; Xneg = DataTrain.B; c = FunPara.c; mu = FunPara.mu; kerfPara = FunPara.kerfPara; m1=size(Xpos,1);m2=size(Xneg,1); e1=ones(m1,1);e2=ones(m2,1); TrainX=[Xpos;Xneg]; [m,n]=size(TestX); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Compute Kernel %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% if ~strcmp(kerfPara.type,'lin') Xpos=kernelfun(Xpos,kerfPara,TrainX); Xneg=kernelfun(Xneg,kerfPara,TrainX); TestX=kernelfun(TestX,kerfPara,TrainX); end Hpos=[Xpos e1]'*[Xpos e1]; Hneg=[Xneg e2]'*[Xneg e2]; M=Hpos-mu*Hneg; N=Hneg-mu*Hpos; if ~strcmp(kerfPara.type,'lin') M=M+c*eye(m1+m2+1); N=N+c*eye(m1+m2+1); else M=M+c*eye(n+1); N=N+c*eye(n+1); end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %Compute (w1,b1) and (w2,b2) %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% [a1,a2]=eig(M); [a3,a4]=eig(N); [~,index_w1]=min(diag(a2)); [~,index_w2]=min(diag(a4)); if strcmp(kerfPara.type,'lin') w1=a1(:,index_w1); w2=-a3(:,index_w2); else w1=a1(1:(m1+m2+1),index_w1); w2=-a3(:,index_w2); end normw1=norm(w1); normw2=norm(w2); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % If necessary; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% if abs(normw1-normw2)>eps a2(:,index_w1)=a2(:,index_w1)+10000; a4(:,index_w2)=a2(:,index_w2)+10000; [~,index_w1]=min(diag(a2)); [~,index_w2]=min(diag(a4)); w11=a1(:,index_w1); w22=-a3(:,index_w2); diff=[]; lamda=0.1; for i=1:1:10 w111=lamda*i*w1+(1-lamda*i)*w11; w222=lamda*i*w2+(1-lamda*i)*w22; normw1=norm(w111); normw2=norm(w222); diff(i)=abs(normw1-normw2); end [~,index_diff]=min(diff); w1=lamda*index_diff*w1+(1-lamda*index_diff)*w11; w2=lamda*index_diff*w2+(1-lamda*index_diff)*w22; end %toc; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Predict and output %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% if strcmp(kerfPara.type,'lin') Y1=TestX*w1(1:n)+w1(n+1)*ones(m,1); %testX Y2=TestX*w2(1:n)+w2(n+1)*ones(m,1); %testX else mm = m1+m2; Y1=TestX*w1(1:mm)+w1(mm+1)*ones(m,1); %testX Y2=TestX*w2(1:mm)+w2(mm+1)*ones(m,1); %testX end Predict_Y = sign(abs(Y2)-abs(Y1));

PCC

Reference

Main Function

Contacts