PLSR

Partial Least Squares Regression

   W = PLSR
   W = PLSR([],MAXLV,METHOD)

   [W, INFORM] = PLSR(A,MAXLV,METHOD)

Input
A	training dataset
MAXLV	maximal number of latent variables (will be corrected  if > rank(A)); MAXLV=inf means MAXLV=min(size(A)) -- theoretical  maximum number of LV; by default = inf
METHOD	'NIPALS' or 'SIMPLS'; by default = 'SIMPLS'

Output
W	PLS feature extraction mapping
INFORM	extra algorithm output

  DESRIPTION PRTools Adaptation of PLS_TRAIN/PLS_APPLY routines. No preprocessing  is done inside this mapping. It is the user responsibility to train  preprocessing on training data and apply it to the test data.

Crisp labels will be converted into soft labels which will be used as  a target matrix.

In order to do regression with the smaller number of latent variables  than the number of LV's selected during trainig do  d = w.data;  d.n = new_n;  w.data = d;

See also

pls_train, pls_transform, pls_apply,