source: trunk/test/distance_test.cc @ 1437

// $Id: distance_test.cc 1437 2008-08-25 17:55:00Z peter $

/*
  Copyright (C) 2007, 2008 Jari Häkkinen, Peter Johansson, Markus Ringnér

  This file is part of the yat library, http://dev.thep.lu.se/yat

  The yat library is free software; you can redistribute it and/or
  modify it under the terms of the GNU General Public License as
  published by the Free Software Foundation; either version 2 of the
  License, or (at your option) any later version.

  The yat library is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  General Public License for more details.

  You should have received a copy of the GNU General Public License
  along with this program; if not, write to the Free Software
  Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
  02111-1307, USA.
*/

#include "Suite.h"

#include "yat/classifier/DataLookupWeighted1D.h"
#include "yat/classifier/MatrixLookupWeighted.h"
#include "yat/statistics/EuclideanDistance.h"
#include "yat/statistics/PearsonDistance.h"
#include "yat/utility/Matrix.h"
#include "yat/utility/Vector.h"

#include <cassert>
#include <fstream>
#include <iostream>
#include <limits>
#include <list>
#include <vector>


using namespace theplu::yat;

void check_equality(double, double, test::Suite&, const std::string&, 
                    unsigned long int N=1);
utility::Matrix data(void);

template<class Distance>
void test_distance(Distance, theplu::yat::test::Suite&, unsigned long int N=1);

template<class Distance>
void test_duplicate(Distance, theplu::yat::test::Suite&, unsigned long int N=1);

template<class Distance>
void test_rescaling(Distance, theplu::yat::test::Suite&, unsigned long int N=1);

template<class Distance>
void test_unity_weights(Distance, theplu::yat::test::Suite&, 
                        unsigned long int N=1);

template<class Distance>
void test_self_distance(Distance, theplu::yat::test::Suite&, 
                        unsigned long int N=1);

template<class Distance>
void test_symmetry(Distance, theplu::yat::test::Suite&, unsigned long int N=1);

template<class Distance>
void test_zero_weight(Distance, theplu::yat::test::Suite&, 
                      unsigned long int N=1);

utility::Matrix weight(void);

int main(int argc, char* argv[])
{  
  theplu::yat::test::Suite suite(argc, argv);
  suite.err() << "testing distance" << std::endl;
  
  utility::Vector a(3,1);
  a(1) = 2;
  utility::Vector b(3,0);
  b(2) = 1;
  
  double tolerance=1e-4;
  statistics::EuclideanDistance eucl_dist;
  suite.err() << "testing EuclideanDistance" << std::endl;
  test_distance(eucl_dist, suite, 100);
  double dist=eucl_dist(a.begin(),a.end(),b.begin());
  if(std::abs(dist-2.23607)>tolerance) {
    suite.err() << "Error in unweighted Euclidean distance " << std::endl;
    suite.add(false);
  }
  
  statistics::PearsonDistance pear_dist;
  suite.err() << "testing PearsonDistance" << std::endl;
  test_distance(pear_dist, suite, 1000);
  dist=pear_dist(a.begin(),a.end(),b.begin());
  if(std::abs(dist-1.5)>tolerance) {
    suite.err() << "Error in unweighted Pearson distance " << std::endl;
    suite.add(false);
  }
  
  
  // Testing weighted versions
  utility::Matrix m(2,3,1);
  m(0,1)=2;
  m(1,0)=0;
  m(1,1)=0;
  utility::Matrix w(2,3,1);
  w(0,0)=0;
  classifier::MatrixLookupWeighted mw(m,w);
  classifier::DataLookupWeighted1D aw(mw,0,true);
  classifier::DataLookupWeighted1D bw(mw,1,true);
  
  dist=eucl_dist(aw.begin(),aw.end(),bw.begin());
  
  if(std::abs(dist-sqrt(6))>tolerance) {
    suite.err() << "Error in weighted Euclidean distance " << std::endl;
    suite.add(false);
  }
  
  dist=pear_dist(aw.begin(),aw.end(),bw.begin());
  
  if(std::abs(dist-2)>tolerance) {
    suite.err() << "Error in weighted Pearson distance " << std::endl;
    suite.add(false);
  }
  
    
  // Test with std::vectors
  std::vector<double> sa(3,1);
  sa[1] = 2;
  std::vector<double> sb(3,0);
  sb[2] = 1;
  
  dist=eucl_dist(sa.begin(),sa.end(),sb.begin()); 
  if(std::abs(dist-2.23607)>tolerance) {
    suite.err() << "Error in distance for std::vector " << std::endl;
    suite.add(false);
  }
  
  // Test for a std::list and a std::vector
  std::list<double> la;
  std::copy(sa.begin(),sa.end(),std::back_inserter<std::list<double> >(la));
  dist=eucl_dist(la.begin(),la.end(),sb.begin());
  if(std::abs(dist-2.23607)>tolerance) {
    suite.err() << "Error in distance for std::list " << std::endl;
    suite.add(false);
  }
  

  return suite.return_value();
}


void check_equality(double dist1, double dist2, test::Suite& suite, 
                    const std::string& msg, unsigned long int N)
{
  if (!suite.equal(dist1, dist2, N)) {
    suite.err() << "Error: " << msg << " failed.\n";
    suite.add(false);
  }
}


utility::Matrix data(void)
{
  utility::Matrix res(2,10);
  for (size_t i = 0; i<res.columns(); ++i){
    res(0,i) = i*i+1;
    res(1,i) = 2*i+3;
  }
  return res;
}

template<class Distance>
void test_distance(Distance dist, theplu::yat::test::Suite& suite,
                   unsigned int long N)
{
  test_duplicate(dist, suite, N);
  test_rescaling(dist, suite, N);
  test_unity_weights(dist, suite, N);
  test_self_distance(dist, suite, N);
  test_symmetry(dist, suite, N);
  test_zero_weight(dist, suite, N);
}

template<class Distance>
void test_duplicate(Distance dist, theplu::yat::test::Suite& suite, 
                    unsigned long int N)
{
  utility::Matrix x(data());
  utility::Matrix x2(x.rows(), 2*x.columns());
  for (size_t i=0; i<x.rows(); ++i){
    std::copy(x.begin_row(i), x.end_row(i), x2.begin_row(i));
    std::copy(x.begin_row(i), x.end_row(i), x2.begin_row(i)+x.columns());
  }
  utility::Matrix w(weight());
  utility::Matrix w2(w.rows(), 2*w.columns());
  for (size_t i=0; i<w.rows(); ++i){
    std::copy(w.begin_row(i), w.end_row(i), w2.begin_row(i));
    std::copy(w.begin_row(i), w.end_row(i), w2.begin_row(i)+w.columns());
  }
  classifier::MatrixLookupWeighted ml(x2, w2);
  double dist1 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  for (size_t i=0; i<w.columns(); ++i)
    w2(0,i)=0.0;
  double dist2 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  check_equality(dist1, dist2, suite, "duplicate property", N);
}

template<class Distance>
void test_rescaling(Distance dist, theplu::yat::test::Suite& suite, 
                    unsigned long int N)
{
  utility::Matrix x=data();
  utility::Matrix w=weight();
  classifier::MatrixLookupWeighted ml(x,w);
  double dist1 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  w *= 2.13;
  double dist2 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  check_equality(dist1, dist2, suite, "rescaling", N);
}

template<class Distance>
void test_unity_weights(Distance dist, theplu::yat::test::Suite& suite, 
                        unsigned long int N)
{
  utility::Matrix x=data();
  classifier::MatrixLookupWeighted ml(x);
  double dist1 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  double dist2 = dist(x.begin_row(0), x.end_row(0), x.begin_row(1));
  check_equality(dist1, dist2, suite, "unity weights", N);
}

template<class Distance>
void test_self_distance(Distance dist, theplu::yat::test::Suite& suite, 
                        unsigned long int N)
{
  utility::Matrix x = data();
  double self = dist(x.begin(), x.end(), x.begin());
  if (!(std::abs(self) <= N*std::numeric_limits<double>().epsilon()) ){
    suite.err() << "error: self distance is " << self << "\n"
                << "supposed to be zero.\n";
    suite.add(false);
  }
}


template<class Distance>
void test_symmetry(Distance dist, theplu::yat::test::Suite& suite, 
                   unsigned long int N)
{
  utility::Matrix x = data();
  double distab = dist(x.begin_row(0), x.end_row(0), x.begin_row(1));
  double distba = dist(x.begin_row(1), x.end_row(1), x.begin_row(0));
  check_equality(distab, distba, suite, "symmetry test", N);
}


template<class Distance>
void test_zero_weight(Distance dist, theplu::yat::test::Suite& suite, 
                      unsigned long int N)
{
  utility::Matrix x=data();
  utility::Matrix w=weight();
  classifier::MatrixLookupWeighted ml(x,w);
  w(0,0) = 0.0;
  double dist1 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  w(0,0) = 100*std::numeric_limits<double>().epsilon();
  double dist2 = dist(ml.begin_row(0), ml.end_row(0), ml.begin_row(1));
  check_equality(dist1, dist2, suite, "zero weight", N);
}

utility::Matrix weight(void)
{
  utility::Matrix res(2,10);
  for (size_t i = 0; i<res.columns(); ++i){
    res(0,i) = 1.0/(1+i);
    res(1,i) = 1.0-0.1*i;
  }
  return res;
}