source: trunk/test/normalization_test.cc @ 1737

Last change on this file since 1737 was 1737, checked in by Peter, 12 years ago

addresses #478 - preparing for weighted version

  • Property svn:eol-style set to native
  • Property svn:keywords set to Id
File size: 10.4 KB
Line 
1// $Id: normalization_test.cc 1737 2009-01-19 20:40:06Z peter $
2
3/*
4  Copyright (C) 2008, 2009 Jari Häkkinen, Peter Johansson
5
6  This file is part of the yat library, http://dev.thep.lu.se/yat
7
8  The yat library is free software; you can redistribute it and/or
9  modify it under the terms of the GNU General Public License as
10  published by the Free Software Foundation; either version 3 of the
11  License, or (at your option) any later version.
12
13  The yat library is distributed in the hope that it will be useful,
14  but WITHOUT ANY WARRANTY; without even the implied warranty of
15  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16  General Public License for more details.
17
18  You should have received a copy of the GNU General Public License
19  along with yat. If not, see <http://www.gnu.org/licenses/>.
20*/
21
22#include "Suite.h"
23
24#include "yat/normalizer/Centralizer.h"
25#include "yat/normalizer/ColumnNormalizer.h"
26#include "yat/normalizer/Gauss.h"
27#include "yat/normalizer/qQuantileNormalizer.h"
28#include "yat/normalizer/QuantileNormalizer.h"
29#include "yat/normalizer/RowNormalizer.h"
30#include "yat/normalizer/Spearman.h"
31#include "yat/normalizer/Zscore.h"
32
33#include "yat/utility/DataIterator.h"
34#include "yat/utility/FileUtil.h"
35#include "yat/utility/Matrix.h"
36#include "yat/utility/MatrixWeighted.h"
37#include "yat/utility/WeightIterator.h"
38
39#include <fstream>
40#include <limits>
41#include <vector>
42
43using namespace theplu::yat;
44void test_centralizer(test::Suite&);
45void test_column_normalize(test::Suite&);
46void test_gauss_normalize(test::Suite&);
47void test_qquantile_normalize(test::Suite&);
48void test_qquantile_normalize_weighted(test::Suite&);
49void test_quantile_normalize(test::Suite&);
50void test_row_normalize(test::Suite&);
51void test_spearman(test::Suite&);
52void test_spearman_weighted(test::Suite&);
53void test_z_score(test::Suite&);
54
55int main(int argc, char* argv[])
56{ 
57  test::Suite suite(argc, argv);
58  suite.err() << "testing normalizations ... " << std::endl;
59
60  test_centralizer(suite);
61  test_column_normalize(suite);
62  test_qquantile_normalize(suite);
63  test_qquantile_normalize_weighted(suite);
64  test_quantile_normalize(suite);
65  test_gauss_normalize(suite);
66  test_row_normalize(suite);
67  test_spearman(suite);
68  test_z_score(suite);
69
70  return suite.return_value();
71}
72
73
74void test_centralizer(test::Suite& suite)
75{
76  suite.err() << "Testing Centralizer\n";
77  std::vector<double> vec;
78  vec.push_back(1);
79  vec.push_back(2);
80  vec.push_back(3);
81  normalizer::Centralizer<> c;
82  c(vec.begin(), vec.end(), vec.begin());
83  for (size_t i=0; i<vec.size(); ++i)
84    suite.add(suite.equal(vec[i], static_cast<double>(i)-1.0));
85
86  std::vector<utility::DataWeight> vec2;
87  vec2.push_back(utility::DataWeight(1,1));
88  vec2.push_back(utility::DataWeight(2,0.5));
89  vec2.push_back(utility::DataWeight(2,0.5));
90  std::vector<utility::DataWeight> vec3(vec2.size());
91  c(vec2.begin(), vec2.end(), vec3.begin());
92  for (size_t i=0; i<vec2.size(); ++i)
93    suite.add(suite.equal(vec3[i].weight(), vec2[i].weight()));
94  suite.add(suite.equal(vec3[0].data(), -0.5));
95  suite.add(suite.equal(vec3[1].data(), 0.5));
96  suite.add(suite.equal(vec3[2].data(), 0.5));
97
98}
99
100
101void test_column_normalize(test::Suite& suite)
102{
103  using namespace normalizer;
104  suite.err() << "Testing ColumnNormalizer\n";
105 
106  utility::Matrix m(2,2);
107  m(0,0) = 0;
108  m(0,1) = 10;
109  m(1,0) = 2;
110  m(1,1) = 4;
111  ColumnNormalizer<Centralizer<> > qn;
112  qn(m, m);
113  suite.err() << "Testing m(0,0)\n";
114  suite.add(suite.equal(m(0,0), -1));
115  suite.err() << "Testing m(0,1)\n";
116  suite.add(suite.equal(m(0,1), 3));
117  suite.err() << "Testing m(1,0)\n";
118  suite.add(suite.equal(m(1,0), 1));
119  suite.err() << "Testing m(1,1)\n";
120  suite.add(suite.equal(m(1,1), -3));
121}
122
123
124void test_qquantile_normalize(test::Suite& suite)
125{
126  using namespace normalizer;
127
128  suite.err() << "Testing qQuantileNormalizer\n";
129  std::string data(test::filename("data/normalization_test.data"));
130  if (utility::FileUtil(data.c_str()).permissions("r")) {
131    suite.add(false);
132    suite.err() << "Cannot access file " << data << '\n';
133    return;
134  }
135  std::ifstream data_stream(data.c_str());
136
137  utility::Matrix m(data_stream);
138
139  suite.err() << "testing number of parts (Q) boundary conditions\n";
140  qQuantileNormalizer(m.begin_column(0), m.end_column(0), m.rows());
141  qQuantileNormalizer(m.begin_column(0), m.end_column(0), 3);
142
143  // first column as target
144  qQuantileNormalizer qqn(m.begin_column(0), m.end_column(0) ,9); 
145  ColumnNormalizer<qQuantileNormalizer> cn(qqn);
146  utility::Matrix result(m.rows(),m.columns());
147  cn(m, result);
148
149  suite.err() << "test that result can be stored in the source matrix\n";
150  cn(m,m);
151  suite.add(result==m);
152
153  // Enough iteration will make all columns to have the same values as
154  // the target.
155  suite.err() << "Testing that q=matrix rows gives QuantileNormaliztion\n";
156  utility::Matrix m2(4,2);
157  m2(0,0) = 0; m2(0,1) = 10;
158  m2(1,0) = 2; m2(1,1) = 4;
159  m2(2,0) = 1; m2(2,1) = 0;
160  m2(3,0) = 3; m2(3,1) = 7;
161  qQuantileNormalizer qqn2(m2.begin_column(0), m2.end_column(0), m2.rows());
162  ColumnNormalizer<qQuantileNormalizer> cn2(qqn2);
163  utility::Matrix result2(m2.rows(),m2.columns());
164  cn2(m2,result2);
165  suite.add( suite.equal_fix(m2(0,0),result2(2,1),1.0e-12) &&
166             suite.equal_fix(m2(1,0),result2(3,1),1.0e-12) &&
167             suite.equal_fix(m2(2,0),result2(1,1),1.0e-12) &&
168             suite.equal_fix(m2(3,0),result2(0,1),1.0e-12) );
169}
170
171
172void test_qquantile_normalize_weighted(test::Suite& suite)
173{
174  using namespace normalizer;
175
176  suite.err() << "Testing qQuantileNormalizer weighted\n";
177
178  std::vector<utility::DataWeight> vec;
179  vec.reserve(100);
180  for (size_t i=0; i<100; ++i)
181    vec.push_back(utility::DataWeight(i, 1.0));
182  //qQuantileNormalizer(vec.begin(), vec.end(), 5);
183}
184
185
186void test_quantile_normalize(test::Suite& suite)
187{
188  suite.err() << "Testing QuantileNormalizer\n";
189 
190  utility::Matrix m(2,2);
191  m(0,0) = 0;
192  m(0,1) = 10;
193  m(1,0) = 2;
194  m(1,1) = 4;
195  normalizer::QuantileNormalizer qn;
196  qn(m, m);
197  suite.err() << "Testing m(0,0)\n";
198  suite.add(suite.equal(m(0,0), 2));
199  suite.err() << "Testing m(0,1)\n";
200  suite.add(suite.equal(m(0,1), 6));
201  suite.err() << "Testing m(1,0)\n";
202  suite.add(suite.equal(m(1,0), 6));
203  suite.err() << "Testing m(1,1)\n";
204  suite.add(suite.equal(m(1,1), 2));
205}
206
207void test_row_normalize(test::Suite& suite)
208{
209  using namespace normalizer;
210  suite.err() << "Testing RowNormalizer\n";
211 
212  utility::Matrix m(2,3);
213  m(0,0) = 0;
214  m(0,1) = 10;
215  m(1,0) = 2;
216  m(1,1) = 4;
217  utility::Matrix m2(m);
218  m2.transpose();
219  ColumnNormalizer<Centralizer<> > cn;
220  RowNormalizer<Centralizer<> > rn;
221  cn(m, m);
222  rn(m2, m2);
223  m2.transpose();
224  suite.equal_range(m.begin(), m.end(), m2.begin());
225}
226
227void test_spearman(test::Suite& suite)
228{
229  suite.err() << "Testing Spearman\n";
230  normalizer::Spearman spearman;
231  std::vector<double> vec;
232  vec.push_back(0);
233  vec.push_back(2);
234  vec.push_back(3);
235  vec.push_back(1);
236  spearman(vec.begin(), vec.end(), vec.begin());
237  std::vector<double> correct;
238  correct.push_back(1.0/8);
239  correct.push_back(5.0/8);
240  correct.push_back(7.0/8);
241  correct.push_back(3.0/8);
242  suite.add(suite.equal_range(vec.begin(), vec.end(), correct.begin()));
243  suite.err() << "Testing Spearman with ties\n";
244  vec[1]=vec[2];
245  correct[1] = correct[2] = (correct[1]+correct[2])/2;
246  spearman(vec.begin(), vec.end(), vec.begin());
247  suite.add(suite.equal_range(vec.begin(), vec.end(), correct.begin()));
248  test_spearman_weighted(suite);
249}
250
251
252void test_gauss_normalize(test::Suite& suite)
253{
254  suite.err() << "Testing Gauss\n";
255  normalizer::Gauss gauss;
256  std::vector<double> vec;
257  vec.push_back(1);
258  gauss(vec.begin(), vec.end(), vec.begin());
259  suite.add(suite.equal(vec.front(), 0));
260  vec.push_back(1);
261  gauss(vec.begin(), vec.end(), vec.begin());
262  suite.add(suite.equal(vec.front(), -vec.back()));
263
264}
265
266void test_spearman_weighted(test::Suite& suite)
267{
268  suite.err() << "Testing Weighted Spearman\n";
269  normalizer::Spearman spearman;
270
271  suite.err() << "Testing that unity weights reproduces unweighted case\n";
272  utility::MatrixWeighted m(1,4,0,1);
273  utility::MatrixWeighted res(m.rows(), m.columns(),3.14,0);
274  m(0,0).data()=0;
275  m(0,1).data()=2;
276  m(0,2).data()=3;
277  m(0,3).data()=1;
278  std::vector<double> correct(m.columns());
279  std::vector<double> correct_w(m.columns(), 1.0);
280  std::copy(utility::data_iterator(m.begin_row(0)),
281            utility::data_iterator(m.end_row(0)),
282            correct.begin());
283  spearman(correct.begin(), correct.end(), correct.begin());
284  spearman(m.begin_row(0), m.end_row(0), res.begin_row(0));
285
286  using utility::data_iterator;
287  suite.add(suite.equal_range(data_iterator(res.begin_row(0)),
288                               data_iterator(res.end_row(0)),
289                               correct.begin()));
290  using utility::weight_iterator;
291  suite.add(suite.equal_range(weight_iterator(res.begin_row(0)),
292                               weight_iterator(res.end_row(0)),
293                               correct_w.begin()));
294
295  suite.err() << "Testing rescaling of weights\n";
296  for (size_t i=0; i<m.columns(); ++i) {
297    m(0,i).weight() *= 2;
298    correct_w[i] *= 2;
299  }   
300  spearman(m.begin_row(0), m.end_row(0), res.begin_row(0));
301  suite.add(suite.equal_range(data_iterator(res.begin_row(0)),
302                               data_iterator(res.end_row(0)),
303                               correct.begin()));
304  suite.add(suite.equal_range(weight_iterator(res.begin_row(0)),
305                               weight_iterator(res.end_row(0)),
306                               correct_w.begin()));
307
308 
309  suite.err() << "Testing case with a zero weight\n";
310  m(0,1).data() = std::numeric_limits<double>::quiet_NaN();
311  m(0,1).weight() = 0.0;
312  spearman(m.begin_row(0), m.end_row(0), res.begin_row(0));
313  suite.add(suite.equal(res(0,0).data(), 0.5/3)); 
314  suite.add(suite.equal(res(0,2).data(), 2.5/3)); 
315  suite.add(suite.equal(res(0,3).data(), 1.5/3)); 
316
317  suite.err() << "Testing case with ties\n";
318  m(0,0).data() = m(0,2).data();
319  spearman(m.begin_row(0), m.end_row(0), res.begin_row(0));
320  suite.add(suite.equal(res(0,0).data(), 2.0/3)); 
321  suite.add(suite.equal(res(0,2).data(), 2.0/3)); 
322  suite.add(suite.equal(res(0,3).data(), 0.5/3)); 
323}
324
325void test_z_score(test::Suite& suite)
326{
327  suite.err() << "Testing Zscore\n";
328  std::vector<double> vec;
329  vec.push_back(0);
330  vec.push_back(3.14);
331  normalizer::Zscore zscore;
332  zscore(vec.begin(), vec.end(), vec.begin());
333  for (size_t i=0; i<vec.size(); ++i)
334    suite.add(suite.equal(vec[i], 2.0*i-1.0));
335
336  std::vector<utility::DataWeight> vec2;
337  vec2.push_back(utility::DataWeight(1,1));
338  vec2.push_back(utility::DataWeight(2.13,0.5));
339  vec2.push_back(utility::DataWeight(2.13,0.5));
340  std::vector<utility::DataWeight> vec3(vec2.size());
341  zscore(vec2.begin(), vec2.end(), vec3.begin());
342  for (size_t i=0; i<vec2.size(); ++i)
343    suite.add(suite.equal(vec3[i].weight(), vec2[i].weight()));
344  suite.add(suite.equal(vec3[0].data(), -1.0));
345  suite.add(suite.equal(vec3[1].data(), 1.0));
346  suite.add(suite.equal(vec3[2].data(), 1.0));
347}
348
349
Note: See TracBrowser for help on using the repository browser.