LIN2SVM

Merges linear rule and kernel projection.



 Synopsis:
  svm_model = lin2svm(kfe_model,lin_model)

 Description:
  This function merges kernel feature extraction model
  (data-type kernel projection) and linear classifier to 
  create kernel (SVM) classifier.

 Input:
  kfe_model [struct] Kernel data projection:
   .Alpha [nsv x new_dim] Weight vector.
   .b [new_dim x 1] Biases.
   .oprions.ker [string] Kernel identifier (see 'help kernel').
   .options.arg [1xnargs] Kernel arguments.

  lin_model [struct] Linear classifier:
   .W [dim x nfun] Weight vector(s).
   .b [nfun x 1] Bias(es).
  
 Output:
  svm_model [struct] Kernel classifer:
   .Alpha [nsv x nfun] Weight vector(s).
   .b [nfun x 1] Bias(es).
   .options [struct] Copy of kfe_model.options.

 Example:
  data = load('riply_trn');
  options = struct('ker','rbf','arg',1,'new_dim',10);
  kpca_model = greedykpca(data.X,options);
  proj_data = kernelproj(data,kpca_model);
  lin_model = fld(proj_data);
  kfd_model = lin2svm(kpca_model,lin_model);
  figure; ppatterns(data); pboundary(kfd_model);

 See also
  LIN2QUADSVMCLASSLINCLASS.


Source: lin2svm.m

(c) Statistical Pattern Recognition Toolbox, (C) 1999-2003,
Written by Vojtech Franc and Vaclav Hlavac,
Czech Technical University Prague,
Faculty of Electrical engineering,
Center for Machine Perception

Modifications:
10-jun-2004, VF
02-Feb-2003, VF