build_classifier.m

libsvm is a simple, easy-to-use, and efficient software for SVM classification and regression. It s

function [CA,SV,b]=Build_classifier(Z_train, C_train,value_lambda,value_gamma)


%     ____________________    INPUTS  ____________________:

%      Z_train       is a matrix of training data, with "N_train" ROWs and "D" COLUMNs 
%                            i^{th} ROW of Z_train is the i^{th} training point
%                                     N_train ---- the number of training data 
%                                           D ---- the dimension of the training data

%      C_train       is a COLUMN vector, with length "N_train" 
%                            C_train(i) is the class label of i^{th} training point 
%                                     C_train(i) = either 1 or -1       !!!!!!


%      value_lambda  is a positive scalar = the value of slack penalty (lambda).



%      value_gamma   is a positive scalar = the value of gamma in RBF kernel, 
%                                 K(u, v) =  exp(-gamma * ( || u-v||^2 ) ).




%   _______________________   OUTPUTS  ____________________:

%       SV      is a matrix of support vectors, with "N_SV" rows and "D" columns
%                       i^{th} row of SV (denoted by sv_i) is the i^{th} support vector
%                                N_SV ------ the number of support vectors
%                                     D ---- defined as above.
%                   

%       CA      is a column vector, with length "N_SV"      
%                       i^{th} element of CA   =   (c_i * alpha_i)   
%                               c_i ----- the original class label of sv_i 
%                                     alpha_i ----- the positive weight associated with sv_i 

%       b       is a scalar = the constant term in the definition of the classifier H(z) 




%  ____________________    How to Use the Outputs ( CA, SV and b ) _______:

%     The SVM classifier H(z) associated with Z_train, C_train, value_lambda
%     and RBF kernel( with parameter value_gamma)  is :

%               H(z) = sign {   [\sum_{i=1}^{N_SV}    c_i * alpha_i * K(sv_i, z) ]    +  b }

%                    = sign {   [\sum_{i=1}^{N_SV}    CA(i) * K(sv_i, z) ]    +  b }

%    ( where
%              N_SV =  number of rows of SV

%              CA(i) is the i^{th} element of CA

%              sv_i  is the i^{th} row of SV, i.e the i^{th} support vector

%              K(sv_i, z) =  exp( -  value_gamma  * ( || sv_i - z ||^2 )     )
%     ).       





fname_train='train.txt';
fname_model='model.txt';

D=size(Z_train,2);

write_data_into_txt(Z_train,C_train,fname_train);

write_cmd(value_lambda, value_gamma, fname_train, fname_model);

svm_cmd;

[CA,SV,b]=readin_model(fname_model,D);