lssvcbay_train.m

The goal of SPID is to provide the user with tools capable to simulate, preprocess, process and clas

function [param, idx_out]=lssvcBay_train(X_train, Y_train, idx_in, varargin)
%function [param, idx_out]=lssvcBay_train(X_train, Y_train, idx_in, varargin)
% Inputs:
% X_train -- Training data matrix of dim (num examples, num features).
% Y_train -- Training output matrix of dim (num examples, 1).
% idx_in -- Indices of the subset of features selected by preprocessing
%           (e.g. [1: size(X_train,2)].)
% varargin -- related parameter settings for lssvm Bayesian training
%            - e.g. {'kernelType', 'rbf', 'maxsteps', -1}
% Returns:
% param -- a structure with two elements
% idx_out -- Indices of the subset of features effectively 
%            used/selected by training 
%           (not needed in this case, as no variable selection is needed).
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

X=X_train(:,idx_in); t=Y_train;
if ~exist('idx_in'), idx_in=1:size(X,2); end;
iniSig=[]; iniMu=1; iniZeta=1; maxsteps=-1;
kernelType='lin';

if nargin>=4
    if length(varargin)==1,
        args=varargin{1};
    else
        args = varargin;
    end
    nargs = length(args);
    if isstr(args{1})
      for i=1:2:nargs
        switch args{i},
         case 'kernelType', kernelType=args{i+1}; 
         case 'maxsteps', maxsteps=args{i+1};
         case 'sigs', iniSig=args{i+1};
         otherwise,
             error(['invalid argument name ' args{i}]);
        end %switch
     end %for
   end %if
end %if nargin

[lssvcB, zmp, zmn, varztrnp, varztrnn ytrn0, zetanew, zetap, zetan alpha, b, sig,gam] = ...
    lssvcmodoutb2_train(X, t, iniMu, iniZeta, kernelType, iniSig, maxsteps, 0);

param=lssvcB; 
idx_out=idx_in;

return