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
LIN2QUAD, SVMCLASS, LINCLASS.
(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