• 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 Gael Guennebaud <gael.guennebaud@inria.fr>
5 // Copyright (C) 2010 Jitse Niesen <jitse@maths.leeds.ac.uk>
6 //
7 // This Source Code Form is subject to the terms of the Mozilla
8 // Public License v. 2.0. If a copy of the MPL was not distributed
9 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
10 
11 #include "main.h"
12 #include <limits>
13 #include <Eigen/Eigenvalues>
14 #include <Eigen/LU>
15 
find_pivot(typename MatrixType::Scalar tol,MatrixType & diffs,Index col=0)16 template<typename MatrixType> bool find_pivot(typename MatrixType::Scalar tol, MatrixType &diffs, Index col=0)
17 {
18   bool match = diffs.diagonal().sum() <= tol;
19   if(match || col==diffs.cols())
20   {
21     return match;
22   }
23   else
24   {
25     Index n = diffs.cols();
26     std::vector<std::pair<Index,Index> > transpositions;
27     for(Index i=col; i<n; ++i)
28     {
29       Index best_index(0);
30       if(diffs.col(col).segment(col,n-i).minCoeff(&best_index) > tol)
31         break;
32 
33       best_index += col;
34 
35       diffs.row(col).swap(diffs.row(best_index));
36       if(find_pivot(tol,diffs,col+1)) return true;
37       diffs.row(col).swap(diffs.row(best_index));
38 
39       // move current pivot to the end
40       diffs.row(n-(i-col)-1).swap(diffs.row(best_index));
41       transpositions.push_back(std::pair<Index,Index>(n-(i-col)-1,best_index));
42     }
43     // restore
44     for(Index k=transpositions.size()-1; k>=0; --k)
45       diffs.row(transpositions[k].first).swap(diffs.row(transpositions[k].second));
46   }
47   return false;
48 }
49 
50 /* Check that two column vectors are approximately equal upto permutations.
51  * Initially, this method checked that the k-th power sums are equal for all k = 1, ..., vec1.rows(),
52  * however this strategy is numerically inacurate because of numerical cancellation issues.
53  */
54 template<typename VectorType>
verify_is_approx_upto_permutation(const VectorType & vec1,const VectorType & vec2)55 void verify_is_approx_upto_permutation(const VectorType& vec1, const VectorType& vec2)
56 {
57   typedef typename VectorType::Scalar Scalar;
58   typedef typename NumTraits<Scalar>::Real RealScalar;
59 
60   VERIFY(vec1.cols() == 1);
61   VERIFY(vec2.cols() == 1);
62   VERIFY(vec1.rows() == vec2.rows());
63 
64   Index n = vec1.rows();
65   RealScalar tol = test_precision<RealScalar>()*test_precision<RealScalar>()*numext::maxi(vec1.squaredNorm(),vec2.squaredNorm());
66   Matrix<RealScalar,Dynamic,Dynamic> diffs = (vec1.rowwise().replicate(n) - vec2.rowwise().replicate(n).transpose()).cwiseAbs2();
67 
68   VERIFY( find_pivot(tol, diffs) );
69 }
70 
71 
eigensolver(const MatrixType & m)72 template<typename MatrixType> void eigensolver(const MatrixType& m)
73 {
74   typedef typename MatrixType::Index Index;
75   /* this test covers the following files:
76      ComplexEigenSolver.h, and indirectly ComplexSchur.h
77   */
78   Index rows = m.rows();
79   Index cols = m.cols();
80 
81   typedef typename MatrixType::Scalar Scalar;
82   typedef typename NumTraits<Scalar>::Real RealScalar;
83 
84   MatrixType a = MatrixType::Random(rows,cols);
85   MatrixType symmA =  a.adjoint() * a;
86 
87   ComplexEigenSolver<MatrixType> ei0(symmA);
88   VERIFY_IS_EQUAL(ei0.info(), Success);
89   VERIFY_IS_APPROX(symmA * ei0.eigenvectors(), ei0.eigenvectors() * ei0.eigenvalues().asDiagonal());
90 
91   ComplexEigenSolver<MatrixType> ei1(a);
92   VERIFY_IS_EQUAL(ei1.info(), Success);
93   VERIFY_IS_APPROX(a * ei1.eigenvectors(), ei1.eigenvectors() * ei1.eigenvalues().asDiagonal());
94   // Note: If MatrixType is real then a.eigenvalues() uses EigenSolver and thus
95   // another algorithm so results may differ slightly
96   verify_is_approx_upto_permutation(a.eigenvalues(), ei1.eigenvalues());
97 
98   ComplexEigenSolver<MatrixType> ei2;
99   ei2.setMaxIterations(ComplexSchur<MatrixType>::m_maxIterationsPerRow * rows).compute(a);
100   VERIFY_IS_EQUAL(ei2.info(), Success);
101   VERIFY_IS_EQUAL(ei2.eigenvectors(), ei1.eigenvectors());
102   VERIFY_IS_EQUAL(ei2.eigenvalues(), ei1.eigenvalues());
103   if (rows > 2) {
104     ei2.setMaxIterations(1).compute(a);
105     VERIFY_IS_EQUAL(ei2.info(), NoConvergence);
106     VERIFY_IS_EQUAL(ei2.getMaxIterations(), 1);
107   }
108 
109   ComplexEigenSolver<MatrixType> eiNoEivecs(a, false);
110   VERIFY_IS_EQUAL(eiNoEivecs.info(), Success);
111   VERIFY_IS_APPROX(ei1.eigenvalues(), eiNoEivecs.eigenvalues());
112 
113   // Regression test for issue #66
114   MatrixType z = MatrixType::Zero(rows,cols);
115   ComplexEigenSolver<MatrixType> eiz(z);
116   VERIFY((eiz.eigenvalues().cwiseEqual(0)).all());
117 
118   MatrixType id = MatrixType::Identity(rows, cols);
119   VERIFY_IS_APPROX(id.operatorNorm(), RealScalar(1));
120 
121   if (rows > 1 && rows < 20)
122   {
123     // Test matrix with NaN
124     a(0,0) = std::numeric_limits<typename MatrixType::RealScalar>::quiet_NaN();
125     ComplexEigenSolver<MatrixType> eiNaN(a);
126     VERIFY_IS_EQUAL(eiNaN.info(), NoConvergence);
127   }
128 
129   // regression test for bug 1098
130   {
131     ComplexEigenSolver<MatrixType> eig(a.adjoint() * a);
132     eig.compute(a.adjoint() * a);
133   }
134 
135   // regression test for bug 478
136   {
137     a.setZero();
138     ComplexEigenSolver<MatrixType> ei3(a);
139     VERIFY_IS_EQUAL(ei3.info(), Success);
140     VERIFY_IS_MUCH_SMALLER_THAN(ei3.eigenvalues().norm(),RealScalar(1));
141     VERIFY((ei3.eigenvectors().transpose()*ei3.eigenvectors().transpose()).eval().isIdentity());
142   }
143 }
144 
eigensolver_verify_assert(const MatrixType & m)145 template<typename MatrixType> void eigensolver_verify_assert(const MatrixType& m)
146 {
147   ComplexEigenSolver<MatrixType> eig;
148   VERIFY_RAISES_ASSERT(eig.eigenvectors());
149   VERIFY_RAISES_ASSERT(eig.eigenvalues());
150 
151   MatrixType a = MatrixType::Random(m.rows(),m.cols());
152   eig.compute(a, false);
153   VERIFY_RAISES_ASSERT(eig.eigenvectors());
154 }
155 
test_eigensolver_complex()156 void test_eigensolver_complex()
157 {
158   int s = 0;
159   for(int i = 0; i < g_repeat; i++) {
160     CALL_SUBTEST_1( eigensolver(Matrix4cf()) );
161     s = internal::random<int>(1,EIGEN_TEST_MAX_SIZE/4);
162     CALL_SUBTEST_2( eigensolver(MatrixXcd(s,s)) );
163     CALL_SUBTEST_3( eigensolver(Matrix<std::complex<float>, 1, 1>()) );
164     CALL_SUBTEST_4( eigensolver(Matrix3f()) );
165     TEST_SET_BUT_UNUSED_VARIABLE(s)
166   }
167   CALL_SUBTEST_1( eigensolver_verify_assert(Matrix4cf()) );
168   s = internal::random<int>(1,EIGEN_TEST_MAX_SIZE/4);
169   CALL_SUBTEST_2( eigensolver_verify_assert(MatrixXcd(s,s)) );
170   CALL_SUBTEST_3( eigensolver_verify_assert(Matrix<std::complex<float>, 1, 1>()) );
171   CALL_SUBTEST_4( eigensolver_verify_assert(Matrix3f()) );
172 
173   // Test problem size constructors
174   CALL_SUBTEST_5(ComplexEigenSolver<MatrixXf> tmp(s));
175 
176   TEST_SET_BUT_UNUSED_VARIABLE(s)
177 }
178