1 // This file is part of Eigen, a lightweight C++ template library
2 // for linear algebra. Eigen itself is part of the KDE project.
3 //
4 // Copyright (C) 2008 Gael Guennebaud <g.gael@free.fr>
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 #define EIGEN_NO_ASSERTION_CHECKING
11 #include "main.h"
12 #include <Eigen/Cholesky>
13 #include <Eigen/LU>
14 
15 #ifdef HAS_GSL
16 #include "gsl_helper.h"
17 #endif
18 
cholesky(const MatrixType & m)19 template<typename MatrixType> void cholesky(const MatrixType& m)
20 {
21   /* this test covers the following files:
22      LLT.h LDLT.h
23   */
24   int rows = m.rows();
25   int cols = m.cols();
26 
27   typedef typename MatrixType::Scalar Scalar;
28   typedef typename NumTraits<Scalar>::Real RealScalar;
29   typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, MatrixType::RowsAtCompileTime> SquareMatrixType;
30   typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, 1> VectorType;
31 
32   MatrixType a0 = MatrixType::Random(rows,cols);
33   VectorType vecB = VectorType::Random(rows), vecX(rows);
34   MatrixType matB = MatrixType::Random(rows,cols), matX(rows,cols);
35   SquareMatrixType symm =  a0 * a0.adjoint();
36   // let's make sure the matrix is not singular or near singular
37   MatrixType a1 = MatrixType::Random(rows,cols);
38   symm += a1 * a1.adjoint();
39 
40   #ifdef HAS_GSL
41   if (ei_is_same_type<RealScalar,double>::ret)
42   {
43     typedef GslTraits<Scalar> Gsl;
44     typename Gsl::Matrix gMatA=0, gSymm=0;
45     typename Gsl::Vector gVecB=0, gVecX=0;
46     convert<MatrixType>(symm, gSymm);
47     convert<MatrixType>(symm, gMatA);
48     convert<VectorType>(vecB, gVecB);
49     convert<VectorType>(vecB, gVecX);
50     Gsl::cholesky(gMatA);
51     Gsl::cholesky_solve(gMatA, gVecB, gVecX);
52     VectorType vecX(rows), _vecX, _vecB;
53     convert(gVecX, _vecX);
54     symm.llt().solve(vecB, &vecX);
55     Gsl::prod(gSymm, gVecX, gVecB);
56     convert(gVecB, _vecB);
57     // test gsl itself !
58     VERIFY_IS_APPROX(vecB, _vecB);
59     VERIFY_IS_APPROX(vecX, _vecX);
60 
61     Gsl::free(gMatA);
62     Gsl::free(gSymm);
63     Gsl::free(gVecB);
64     Gsl::free(gVecX);
65   }
66   #endif
67 
68   {
69     LDLT<SquareMatrixType> ldlt(symm);
70     VERIFY(ldlt.isPositiveDefinite());
71     // in eigen3, LDLT is pivoting
72     //VERIFY_IS_APPROX(symm, ldlt.matrixL() * ldlt.vectorD().asDiagonal() * ldlt.matrixL().adjoint());
73     ldlt.solve(vecB, &vecX);
74     VERIFY_IS_APPROX(symm * vecX, vecB);
75     ldlt.solve(matB, &matX);
76     VERIFY_IS_APPROX(symm * matX, matB);
77   }
78 
79   {
80     LLT<SquareMatrixType> chol(symm);
81     VERIFY(chol.isPositiveDefinite());
82     VERIFY_IS_APPROX(symm, chol.matrixL() * chol.matrixL().adjoint());
83     chol.solve(vecB, &vecX);
84     VERIFY_IS_APPROX(symm * vecX, vecB);
85     chol.solve(matB, &matX);
86     VERIFY_IS_APPROX(symm * matX, matB);
87   }
88 
89 #if 0 // cholesky is not rank-revealing anyway
90   // test isPositiveDefinite on non definite matrix
91   if (rows>4)
92   {
93     SquareMatrixType symm =  a0.block(0,0,rows,cols-4) * a0.block(0,0,rows,cols-4).adjoint();
94     LLT<SquareMatrixType> chol(symm);
95     VERIFY(!chol.isPositiveDefinite());
96     LDLT<SquareMatrixType> cholnosqrt(symm);
97     VERIFY(!cholnosqrt.isPositiveDefinite());
98   }
99 #endif
100 }
101 
test_eigen2_cholesky()102 void test_eigen2_cholesky()
103 {
104   for(int i = 0; i < g_repeat; i++) {
105     CALL_SUBTEST_1( cholesky(Matrix<double,1,1>()) );
106     CALL_SUBTEST_2( cholesky(Matrix2d()) );
107     CALL_SUBTEST_3( cholesky(Matrix3f()) );
108     CALL_SUBTEST_4( cholesky(Matrix4d()) );
109     CALL_SUBTEST_5( cholesky(MatrixXcd(7,7)) );
110     CALL_SUBTEST_6( cholesky(MatrixXf(17,17)) );
111     CALL_SUBTEST_7( cholesky(MatrixXd(33,33)) );
112   }
113 }
114