• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1  // This file is part of Eigen, a lightweight C++ template library
2  // for linear algebra.
3  //
4  // Copyright (C) 2008-2009 Benoit Jacob <jacob.benoit.1@gmail.com>
5  //
6  // This Source Code Form is subject to the terms of the Mozilla
7  // Public License v. 2.0. If a copy of the MPL was not distributed
8  // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
9  
10  #include "main.h"
11  #include <Eigen/LU>
12  using namespace std;
13  
lu_non_invertible()14  template<typename MatrixType> void lu_non_invertible()
15  {
16    typedef typename MatrixType::Index Index;
17    typedef typename MatrixType::RealScalar RealScalar;
18    /* this test covers the following files:
19       LU.h
20    */
21    Index rows, cols, cols2;
22    if(MatrixType::RowsAtCompileTime==Dynamic)
23    {
24      rows = internal::random<Index>(2,EIGEN_TEST_MAX_SIZE);
25    }
26    else
27    {
28      rows = MatrixType::RowsAtCompileTime;
29    }
30    if(MatrixType::ColsAtCompileTime==Dynamic)
31    {
32      cols = internal::random<Index>(2,EIGEN_TEST_MAX_SIZE);
33      cols2 = internal::random<int>(2,EIGEN_TEST_MAX_SIZE);
34    }
35    else
36    {
37      cols2 = cols = MatrixType::ColsAtCompileTime;
38    }
39  
40    enum {
41      RowsAtCompileTime = MatrixType::RowsAtCompileTime,
42      ColsAtCompileTime = MatrixType::ColsAtCompileTime
43    };
44    typedef typename internal::kernel_retval_base<FullPivLU<MatrixType> >::ReturnType KernelMatrixType;
45    typedef typename internal::image_retval_base<FullPivLU<MatrixType> >::ReturnType ImageMatrixType;
46    typedef Matrix<typename MatrixType::Scalar, ColsAtCompileTime, ColsAtCompileTime>
47            CMatrixType;
48    typedef Matrix<typename MatrixType::Scalar, RowsAtCompileTime, RowsAtCompileTime>
49            RMatrixType;
50  
51    Index rank = internal::random<Index>(1, (std::min)(rows, cols)-1);
52  
53    // The image of the zero matrix should consist of a single (zero) column vector
54    VERIFY((MatrixType::Zero(rows,cols).fullPivLu().image(MatrixType::Zero(rows,cols)).cols() == 1));
55  
56    MatrixType m1(rows, cols), m3(rows, cols2);
57    CMatrixType m2(cols, cols2);
58    createRandomPIMatrixOfRank(rank, rows, cols, m1);
59  
60    FullPivLU<MatrixType> lu;
61  
62    // The special value 0.01 below works well in tests. Keep in mind that we're only computing the rank
63    // of singular values are either 0 or 1.
64    // So it's not clear at all that the epsilon should play any role there.
65    lu.setThreshold(RealScalar(0.01));
66    lu.compute(m1);
67  
68    MatrixType u(rows,cols);
69    u = lu.matrixLU().template triangularView<Upper>();
70    RMatrixType l = RMatrixType::Identity(rows,rows);
71    l.block(0,0,rows,(std::min)(rows,cols)).template triangularView<StrictlyLower>()
72      = lu.matrixLU().block(0,0,rows,(std::min)(rows,cols));
73  
74    VERIFY_IS_APPROX(lu.permutationP() * m1 * lu.permutationQ(), l*u);
75  
76    KernelMatrixType m1kernel = lu.kernel();
77    ImageMatrixType m1image = lu.image(m1);
78  
79    VERIFY_IS_APPROX(m1, lu.reconstructedMatrix());
80    VERIFY(rank == lu.rank());
81    VERIFY(cols - lu.rank() == lu.dimensionOfKernel());
82    VERIFY(!lu.isInjective());
83    VERIFY(!lu.isInvertible());
84    VERIFY(!lu.isSurjective());
85    VERIFY((m1 * m1kernel).isMuchSmallerThan(m1));
86    VERIFY(m1image.fullPivLu().rank() == rank);
87    VERIFY_IS_APPROX(m1 * m1.adjoint() * m1image, m1image);
88  
89    m2 = CMatrixType::Random(cols,cols2);
90    m3 = m1*m2;
91    m2 = CMatrixType::Random(cols,cols2);
92    // test that the code, which does resize(), may be applied to an xpr
93    m2.block(0,0,m2.rows(),m2.cols()) = lu.solve(m3);
94    VERIFY_IS_APPROX(m3, m1*m2);
95  }
96  
lu_invertible()97  template<typename MatrixType> void lu_invertible()
98  {
99    /* this test covers the following files:
100       LU.h
101    */
102    typedef typename NumTraits<typename MatrixType::Scalar>::Real RealScalar;
103    DenseIndex size = MatrixType::RowsAtCompileTime;
104    if( size==Dynamic)
105      size = internal::random<DenseIndex>(1,EIGEN_TEST_MAX_SIZE);
106  
107    MatrixType m1(size, size), m2(size, size), m3(size, size);
108    FullPivLU<MatrixType> lu;
109    lu.setThreshold(RealScalar(0.01));
110    do {
111      m1 = MatrixType::Random(size,size);
112      lu.compute(m1);
113    } while(!lu.isInvertible());
114  
115    VERIFY_IS_APPROX(m1, lu.reconstructedMatrix());
116    VERIFY(0 == lu.dimensionOfKernel());
117    VERIFY(lu.kernel().cols() == 1); // the kernel() should consist of a single (zero) column vector
118    VERIFY(size == lu.rank());
119    VERIFY(lu.isInjective());
120    VERIFY(lu.isSurjective());
121    VERIFY(lu.isInvertible());
122    VERIFY(lu.image(m1).fullPivLu().isInvertible());
123    m3 = MatrixType::Random(size,size);
124    m2 = lu.solve(m3);
125    VERIFY_IS_APPROX(m3, m1*m2);
126    VERIFY_IS_APPROX(m2, lu.inverse()*m3);
127  
128    // Regression test for Bug 302
129    MatrixType m4 = MatrixType::Random(size,size);
130    VERIFY_IS_APPROX(lu.solve(m3*m4), lu.solve(m3)*m4);
131  }
132  
lu_partial_piv()133  template<typename MatrixType> void lu_partial_piv()
134  {
135    /* this test covers the following files:
136       PartialPivLU.h
137    */
138    typedef typename MatrixType::Index Index;
139    Index rows = internal::random<Index>(1,4);
140    Index cols = rows;
141  
142    MatrixType m1(cols, rows);
143    m1.setRandom();
144    PartialPivLU<MatrixType> plu(m1);
145  
146    VERIFY_IS_APPROX(m1, plu.reconstructedMatrix());
147  }
148  
lu_verify_assert()149  template<typename MatrixType> void lu_verify_assert()
150  {
151    MatrixType tmp;
152  
153    FullPivLU<MatrixType> lu;
154    VERIFY_RAISES_ASSERT(lu.matrixLU())
155    VERIFY_RAISES_ASSERT(lu.permutationP())
156    VERIFY_RAISES_ASSERT(lu.permutationQ())
157    VERIFY_RAISES_ASSERT(lu.kernel())
158    VERIFY_RAISES_ASSERT(lu.image(tmp))
159    VERIFY_RAISES_ASSERT(lu.solve(tmp))
160    VERIFY_RAISES_ASSERT(lu.determinant())
161    VERIFY_RAISES_ASSERT(lu.rank())
162    VERIFY_RAISES_ASSERT(lu.dimensionOfKernel())
163    VERIFY_RAISES_ASSERT(lu.isInjective())
164    VERIFY_RAISES_ASSERT(lu.isSurjective())
165    VERIFY_RAISES_ASSERT(lu.isInvertible())
166    VERIFY_RAISES_ASSERT(lu.inverse())
167  
168    PartialPivLU<MatrixType> plu;
169    VERIFY_RAISES_ASSERT(plu.matrixLU())
170    VERIFY_RAISES_ASSERT(plu.permutationP())
171    VERIFY_RAISES_ASSERT(plu.solve(tmp))
172    VERIFY_RAISES_ASSERT(plu.determinant())
173    VERIFY_RAISES_ASSERT(plu.inverse())
174  }
175  
test_lu()176  void test_lu()
177  {
178    for(int i = 0; i < g_repeat; i++) {
179      CALL_SUBTEST_1( lu_non_invertible<Matrix3f>() );
180      CALL_SUBTEST_1( lu_invertible<Matrix3f>() );
181      CALL_SUBTEST_1( lu_verify_assert<Matrix3f>() );
182  
183      CALL_SUBTEST_2( (lu_non_invertible<Matrix<double, 4, 6> >()) );
184      CALL_SUBTEST_2( (lu_verify_assert<Matrix<double, 4, 6> >()) );
185  
186      CALL_SUBTEST_3( lu_non_invertible<MatrixXf>() );
187      CALL_SUBTEST_3( lu_invertible<MatrixXf>() );
188      CALL_SUBTEST_3( lu_verify_assert<MatrixXf>() );
189  
190      CALL_SUBTEST_4( lu_non_invertible<MatrixXd>() );
191      CALL_SUBTEST_4( lu_invertible<MatrixXd>() );
192      CALL_SUBTEST_4( lu_partial_piv<MatrixXd>() );
193      CALL_SUBTEST_4( lu_verify_assert<MatrixXd>() );
194  
195      CALL_SUBTEST_5( lu_non_invertible<MatrixXcf>() );
196      CALL_SUBTEST_5( lu_invertible<MatrixXcf>() );
197      CALL_SUBTEST_5( lu_verify_assert<MatrixXcf>() );
198  
199      CALL_SUBTEST_6( lu_non_invertible<MatrixXcd>() );
200      CALL_SUBTEST_6( lu_invertible<MatrixXcd>() );
201      CALL_SUBTEST_6( lu_partial_piv<MatrixXcd>() );
202      CALL_SUBTEST_6( lu_verify_assert<MatrixXcd>() );
203  
204      CALL_SUBTEST_7(( lu_non_invertible<Matrix<float,Dynamic,16> >() ));
205  
206      // Test problem size constructors
207      CALL_SUBTEST_9( PartialPivLU<MatrixXf>(10) );
208      CALL_SUBTEST_9( FullPivLU<MatrixXf>(10, 20); );
209    }
210  }
211