plscvbkf.m

偏最小二乘算法在MATLAB中的实现

function [press,cumpress,minlv,b] = plscvbkf(x,y,split,lv)
%PLSCVBKF Fast Cross validation for PLS using contiguous data blocks
%  This function is primarily intended for use with GENALG
%  Inputs are the matrix of predictor variables (x), vector
%  of predicted variable (y), number of divisions of the data
%  (split),  and maximum number of latent variables to calculate (lv).
%  Outputs are the prediction residual error sum of squares for each 
%  test set (press), cumulative PRESS (cumpress), number of latent 
%  variables at minimum PRESS (minlv), and the final regression vector 
%  (b) at minimum PRESS.  
%  
%  This cross validation routine forms the test sets out of 
%  contiguous blocks of data. Note that this routine does not
%  mean center each test set. The primary emphasis for this routine
%  is speed.
%
%  I/O format is: 
%  [press,cumpress,minlv,b] = plscvbkf(x,y,split,lv);

%  Copyright
%  Eigenvector Technologies
%  1995

[mx,nx] = size(x);
[my,ny] = size(y);
if mx ~= my
  error('Number of samples must be the same in both blocks')
end
press = zeros(split,lv);
ind = ones(split,2);
for i = 1:split
  ind(i,2) = round(i*mx/split);
end 
for i = 1:split-1
  ind(i+1,1) = ind(i,2) +1;
end
for i = 1:split
  calx = [x(1:ind(i,1)-1,:); x(ind(i,2)+1:mx,:)];
  testx = x(ind(i,1):ind(i,2),:);
  caly = [y(1:ind(i,1)-1,:); y(ind(i,2)+1:mx,:)];
  testy = y(ind(i,1):ind(i,2),:);
  bbr = simpls1(calx,caly,lv);
  for j = 1:lv
    ypred = testx*bbr((j-1)*ny+1:j*ny,:)';
    press(i,j) = sum(sum((ypred-testy).^2));
  end
end
cumpress = sum(press);
[a,minlv] = min(cumpress);
if nargout > 3
  b = simpls1(x,y,minlv);
  b = b(minlv,:);
end