function [accuracy] = SDM(sourceFile,targetFile, testFile)

% function [accuracy] = SDM(sourceFile,targetFile, testFile)

% Input:

% sourceFile - source domain training data (e.g. source.arff)

% targetFile - target domain training data (e.g. target.arff)

% testFile - target domain testing data (e.g. test.arff)

%

% Output:

% accuracy - final accuracy

%% Preparing data by reading *.arff files and converting into matlab format

%% Set paths for utils files

addpath('.\utils');

fprintf('SDM starts....\nPreparing data...\n');

[Xs, Ys,target_data, test_data] = prepareData(sourceFile,targetFile, testFile);

nt=size(test_data.test_labels,1);

%% Set parameters for SVM

param.C = 10;

param.Cu = 1; % Cu should be less than C

param.Cu_max = 10*param.Cu; % add at most rho patterns at each iteration

param.rho = 10;

param.max_iter = 100;

param.max_unl_num = 5;

param.kernel_type ='gaussian'; % 'gaussian' or 'linear';

%% Set optiones for MMD and Manifold

dd=20;

[numRows,numCols] = size(Xs);

if dd>numRows

dd=numRows-1;

end

if dd>numCols

dd=numCols-1;

end

[numRows,numCols] = size(target_data.target_features);

if dd>numRows

dd=numRows-1;

end

if dd>numCols

dd=numCols-1;

end

options.d = dd; %defualt 20

options.rho = 1.0;

options.p = 10;

options.lambda = 10.0;

options.eta = 0.1;

options.T = 10;

%% Find number of classes and hypothesis

CV=unique(Ys);

Nc=length(CV);

H=Nc*(Nc-1)/2;

%% 1-to-1 multiclass classification

fprintf('Running multiClassSDM...\n');

[predictions] = multiClassSDM(Xs, Ys,target_data, test_data, param, options);

%% Calculate classification accuracy on the test data

fprintf('Calculating classification accuracy...\n');

[accuracy] = getAccuracy(test_data.test_labels, predictions, H, nt, CV);

end