calculatekery.m

一个简单的pls程序吧

function yy=calculatekery(X,Y,Xt)
[inds, dists] = findneib(X,Xt,50);
X=X(inds,:);Y=Y(inds,:);

% %%%%%%%%%%%%%%%%%%%%%%%%%%kpls
n=size(X,1);nt=size(Xt,1);
mn=mean(Y);
Y=Y-ones(n,1)*mn;

K=Kernel(X,'G',2.5,0);
K_t=Kernel_Test(X,Xt,'G',2.5,0);

%%%% centralization K, K_t, (centralization of Y and Yt already done above)  
M=eye(n)-ones(n,n)/n;
Mt=ones(nt,n)/n;
K_t = (K_t - Mt*K)*M;
K=M*K*M;

%mn=mean(Y);


%%%% number of used latent vectors (componets)   
Fac=6; 
[B,T]=KerNIPALS(K,Y,Fac,0);    %%% a) NIPALS based KPLS  
%[B,T] = KerPLS_eig(K,Y,Fac,0); %%% b) K*Y*Y'*t = a *t based KPLS 
% [B,T,U]=KerSIMPLS1(K,Y,Fac);   %%% c) Kernel SIMPLS for single output (this equals a) and b))   
%%%% prediction (train / test)
%Y_hat=T*(T'*Y);   %%%  this is an alternative way for predictions on training set
% Y_hat=K*B + mn    ;
yy=K_t*B + mn ;
% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% [P_X,P_Xt,W,D]=KPCA(X,Xt,10,'G',1,0); 
% [Y_hat,Yt_hat,B]=KPCR(P_X,P_Xt,D,Y,[]); 
% yy=Yt_hat;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%55