1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 package org.apache.commons.math.random;
19
20 import org.apache.commons.math.DimensionMismatchException;
21 import org.apache.commons.math.linear.RealMatrix;
22 import org.apache.commons.math.linear.RealMatrixImpl;
23 import org.apache.commons.math.stat.descriptive.moment.VectorialCovariance;
24 import org.apache.commons.math.stat.descriptive.moment.VectorialMean;
25
26 import junit.framework.*;
27
28 public class CorrelatedRandomVectorGeneratorTest
29 extends TestCase {
30
31 public CorrelatedRandomVectorGeneratorTest(String name) {
32 super(name);
33 mean = null;
34 covariance = null;
35 generator = null;
36 }
37
38 public void testRank() {
39 assertEquals(3, generator.getRank());
40 }
41
42 public void testRootMatrix() {
43 RealMatrix b = generator.getRootMatrix();
44 RealMatrix bbt = b.multiply(b.transpose());
45 for (int i = 0; i < covariance.getRowDimension(); ++i) {
46 for (int j = 0; j < covariance.getColumnDimension(); ++j) {
47 assertEquals(covariance.getEntry(i, j), bbt.getEntry(i, j), 1.0e-12);
48 }
49 }
50 }
51
52 public void testMeanAndCovariance() throws DimensionMismatchException {
53
54 VectorialMean meanStat = new VectorialMean(mean.length);
55 VectorialCovariance covStat = new VectorialCovariance(mean.length, true);
56 for (int i = 0; i < 5000; ++i) {
57 double[] v = generator.nextVector();
58 meanStat.increment(v);
59 covStat.increment(v);
60 }
61
62 double[] estimatedMean = meanStat.getResult();
63 RealMatrix estimatedCovariance = covStat.getResult();
64 for (int i = 0; i < estimatedMean.length; ++i) {
65 assertEquals(mean[i], estimatedMean[i], 0.07);
66 for (int j = 0; j <= i; ++j) {
67 assertEquals(covariance.getEntry(i, j),
68 estimatedCovariance.getEntry(i, j),
69 0.1 * (1.0 + Math.abs(mean[i])) * (1.0 + Math.abs(mean[j])));
70 }
71 }
72
73 }
74
75 public void setUp() {
76 try {
77 mean = new double[] { 0.0, 1.0, -3.0, 2.3};
78
79 RealMatrixImpl b = new RealMatrixImpl(4, 3);
80 double[][] bData = b.getDataRef();
81 int counter = 0;
82 for (int i = 0; i < bData.length; ++i) {
83 double[] bi = bData[i];
84 for (int j = 0; j < b.getColumnDimension(); ++j) {
85 bi[j] = 1.0 + 0.1 * ++counter;
86 }
87 }
88 RealMatrix bbt = b.multiply(b.transpose());
89 covariance = new RealMatrixImpl(mean.length, mean.length);
90 double[][] covData = covariance.getDataRef();
91 for (int i = 0; i < covariance.getRowDimension(); ++i) {
92 covData[i][i] = bbt.getEntry(i, i);
93 for (int j = 0; j < covariance.getColumnDimension(); ++j) {
94 double s = bbt.getEntry(i, j);
95 covData[i][j] = s;
96 covData[j][i] = s;
97 }
98 }
99
100 RandomGenerator rg = new JDKRandomGenerator();
101 rg.setSeed(17399225432l);
102 GaussianRandomGenerator rawGenerator = new GaussianRandomGenerator(rg);
103 generator = new CorrelatedRandomVectorGenerator(mean,
104 covariance,
105 1.0e-12 * covariance.getNorm(),
106 rawGenerator);
107 } catch (DimensionMismatchException e) {
108 fail(e.getMessage());
109 } catch (NotPositiveDefiniteMatrixException e) {
110 fail("not positive definite matrix");
111 }
112 }
113
114 public void tearDown() {
115 mean = null;
116 covariance = null;
117 generator = null;
118 }
119
120 public static Test suite() {
121 return new TestSuite(CorrelatedRandomVectorGeneratorTest.class);
122 }
123
124 private double[] mean;
125 private RealMatrixImpl covariance;
126 private CorrelatedRandomVectorGenerator generator;
127
128 }