#include <iostream>

#include "svm_context.h"

void xor_test();

int main(int argsc, char** args) {

xor_test();

getchar();

return 0;

}

void xor_test() {

// XOR test

// create the matrix of xor truth table

// inputs:

std::vector<std::vector<double>> attribute(4, std::vector<double>(2));

attribute[0][0] = 0.0;

attribute[0][1] = 0.0;

attribute[1][0] = 0.0;

attribute[1][1] = 1.0;

attribute[2][0] = 1.0;

attribute[2][1] = 0.0;

attribute[3][0] = 1.0;

attribute[3][1] = 1.0;

// outputs:

std::vector<int> label(4);

label[0] = 0;

label[1] = 1;

label[2] = 1;

label[3] = 0;

// initialize svm object with default settings (edit the svm_context::param to change any parameters)

// ideally those settings (especially the kernel parameters) should come from svm training using libsvm library

svm_context _svm;

if (_svm.init(attribute, label)) {

// create svm model using default settings

LIB_SVM::svm_model *model = _svm.generate_model();

//testing the resulting model with testing node

std::vector<double>sample(2);

sample[0] = 1;

sample[1] = 0;

double prediction = 0;

std::vector<double> probability(2);

_svm.predict_probability(sample, prediction, probability);

_svm.predict(sample, prediction);

//display the results to standard output

// libsvm creates probability value for each label and assigns the prediction

// to the label with maximum probability

std::cout

<< "\n***********************************\n"

<< "prediction: " << prediction

<< "\nprobability: " << probability[0] << "\t" << probability[1] << std::endl;

}

else

std::cout << "error in initializing svm" << std::endl;

}