nu_rbfsvc_train.m

CU SVM Classifier Matlab Toolbox

function [output]=nu_RbfSVC_Train(data, savefile, C, gamma, nu, options)
%options: options.crossvalidation=0 if crossvalidation is not needed.
%         otherwise, options.crossvalidation=v(v>=2) for a v-fold validation
%         options.probability=0 if probability estimation is not needed.
%         options.probability=1 for probability estimation.
%         options.shrinking=0 if shrinking is not needed.
%         options.shrinking=1 for using shringking.
%         options.multiclass_type=
%               0 -- one against one\n
%               1 -- one against all\n"
%               2 -- dense\n"
%               3 -- sparse\n"
%               4 -- user defined code matrix\n'
%                    the code matrix is stored in the file 'UserCodeMat.dat'
%                    with the format:
%                    row(number of binary classifier) col(number of class)
%                    row*col code matrix
%         options.nr_weight=0 if no class weight need to be changed.
%         options.nr_weight=1 for using different weight for classes.
%         the class label and corresponding weights are specified in
%         options.weight_label and options.weight.

if (nargin < 2) || (nargin > 6)
   disp(' Incorrect number of input variables.\n');
   help nu_RbfSVC_Train;
   return;
end

if(savefile==0)
    filename='CU_SVModel.dat';
else
    fprintf('Please select the model file name! \n');
    [name,path]=uiputfile('*.*','Save model as');
    if name~=0
        filename=strcat(path,name);
    else
        fprintf('Error open file! \n');
        return;
    end
end

if ~isfield(data, 'X')
    disp(' Error: invalid data structure.\n');
    return; 
end
if ~isfield(data, 'y')
    disp(' Error: invalid data structure.\n');
    return; 
end
if isempty(data.X)
    disp(' Error: sample field is empty.\n');
    return; 
end
if size(data.X,2) ~= size(data.y,2)
    disp(' Error: sample and label do not match.\n');
    return; 
end

[row, col]=size(data.X);
hfile=['CUSVM.H'];
if ~libisloaded('CUSVM')
    loadlibrary('CUSVM',hfile);
end

params=libstruct('svm_parameter');
params.svm_type = 1;
params.kernel_type = 2;
params.degree = 3;
params.gamma = 1/row;
params.coef0 = 0;
params.nu = 0.5;
params.cache_size = 40;
params.C = 1;
params.eps = 1e-3;
params.p = 0.1;
params.shrinking = 1;
params.multiclass_type = 0;
params.probability = 0;
params.nr_weight = 0;

inputs=libstruct('svm_data_input');
inputs.m=row;
inputs.n=col;
inputs.labels=data.y;
inputs.samples=data.X;

crossvalidation=zeros(5,1);

if (nargin == 3)
    params.C =C;
elseif (nargin == 4)
    params.gamma =gamma;
    params.C =C;
elseif (nargin == 5)
    params.gamma =gamma;
    params.C =C;
    params.nu = nu;
elseif (nargin == 6)
    params.gamma =gamma;
    params.C =C;
    params.nu = nu;
    if isfield(options, 'crossvalidation')
        if options.crossvalidation>=2
            crossvalidation(1)=1;
            crossvalidation(2)=options.crossvalidation;
            if  (savefile~=0)
                fprintf('Since you choosed crossvalidation! \n No file will be saved!\n');
            end
        end
    end
    if isfield(options, 'probability')
        params.probability =options.probability;
    end
    if isfield(options, 'shrinking')
        params.shrinking =options.shrinking;
    end
    if isfield(options, 'multiclass_type')
        params.multiclass_type =options.multiclass_type;
    end
    if isfield(options, 'nr_weight')
        if (options.nr_weight~=0)&&(~isempty(options.weight_label))&&(~isempty(options.weight))
            params.nr_weight =options.nr_weight;
            params.weight_label=options.weight_label;
            params.weight=options.weight;
        end
    end
end
[name, crossvalidation]=calllib('CUSVM', 'SVM_Train', inputs,  params, filename, crossvalidation);
clear inputs;
clear params;
%unloadlibrary CUSVM;
if crossvalidation(1)~=0
    output.Accuracy=crossvalidation(5);
    output.MSE=crossvalidation(3);
    output.SCC=crossvalidation(4);
else
    output=Read_Model(filename);
end