cart-elc

Source code for CART-ELC
git clone git://git.laack.co/cart-elc.git
Log | Files | Refs | README | LICENSE

householder.cpp (6286B)


      1 // This file is part of Eigen, a lightweight C++ template library
      2 // for linear algebra.
      3 //
      4 // Copyright (C) 2009-2010 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/QR>
     12 
     13 template<typename MatrixType> void householder(const MatrixType& m)
     14 {
     15   static bool even = true;
     16   even = !even;
     17   /* this test covers the following files:
     18      Householder.h
     19   */
     20   Index rows = m.rows();
     21   Index cols = m.cols();
     22 
     23   typedef typename MatrixType::Scalar Scalar;
     24   typedef typename NumTraits<Scalar>::Real RealScalar;
     25   typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, 1> VectorType;
     26   typedef Matrix<Scalar, internal::decrement_size<MatrixType::RowsAtCompileTime>::ret, 1> EssentialVectorType;
     27   typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, MatrixType::RowsAtCompileTime> SquareMatrixType;
     28   typedef Matrix<Scalar, Dynamic, MatrixType::ColsAtCompileTime> HBlockMatrixType;
     29   typedef Matrix<Scalar, Dynamic, 1> HCoeffsVectorType;
     30 
     31   typedef Matrix<Scalar, MatrixType::ColsAtCompileTime, MatrixType::RowsAtCompileTime> TMatrixType;
     32   
     33   Matrix<Scalar, EIGEN_SIZE_MAX(MatrixType::RowsAtCompileTime,MatrixType::ColsAtCompileTime), 1> _tmp((std::max)(rows,cols));
     34   Scalar* tmp = &_tmp.coeffRef(0,0);
     35 
     36   Scalar beta;
     37   RealScalar alpha;
     38   EssentialVectorType essential;
     39 
     40   VectorType v1 = VectorType::Random(rows), v2;
     41   v2 = v1;
     42   v1.makeHouseholder(essential, beta, alpha);
     43   v1.applyHouseholderOnTheLeft(essential,beta,tmp);
     44   VERIFY_IS_APPROX(v1.norm(), v2.norm());
     45   if(rows>=2) VERIFY_IS_MUCH_SMALLER_THAN(v1.tail(rows-1).norm(), v1.norm());
     46   v1 = VectorType::Random(rows);
     47   v2 = v1;
     48   v1.applyHouseholderOnTheLeft(essential,beta,tmp);
     49   VERIFY_IS_APPROX(v1.norm(), v2.norm());
     50 
     51   // reconstruct householder matrix:
     52   SquareMatrixType id, H1, H2;
     53   id.setIdentity(rows, rows);
     54   H1 = H2 = id;
     55   VectorType vv(rows);
     56   vv << Scalar(1), essential;
     57   H1.applyHouseholderOnTheLeft(essential, beta, tmp);
     58   H2.applyHouseholderOnTheRight(essential, beta, tmp);
     59   VERIFY_IS_APPROX(H1, H2);
     60   VERIFY_IS_APPROX(H1, id - beta * vv*vv.adjoint());
     61 
     62   MatrixType m1(rows, cols),
     63              m2(rows, cols);
     64 
     65   v1 = VectorType::Random(rows);
     66   if(even) v1.tail(rows-1).setZero();
     67   m1.colwise() = v1;
     68   m2 = m1;
     69   m1.col(0).makeHouseholder(essential, beta, alpha);
     70   m1.applyHouseholderOnTheLeft(essential,beta,tmp);
     71   VERIFY_IS_APPROX(m1.norm(), m2.norm());
     72   if(rows>=2) VERIFY_IS_MUCH_SMALLER_THAN(m1.block(1,0,rows-1,cols).norm(), m1.norm());
     73   VERIFY_IS_MUCH_SMALLER_THAN(numext::imag(m1(0,0)), numext::real(m1(0,0)));
     74   VERIFY_IS_APPROX(numext::real(m1(0,0)), alpha);
     75 
     76   v1 = VectorType::Random(rows);
     77   if(even) v1.tail(rows-1).setZero();
     78   SquareMatrixType m3(rows,rows), m4(rows,rows);
     79   m3.rowwise() = v1.transpose();
     80   m4 = m3;
     81   m3.row(0).makeHouseholder(essential, beta, alpha);
     82   m3.applyHouseholderOnTheRight(essential.conjugate(),beta,tmp);
     83   VERIFY_IS_APPROX(m3.norm(), m4.norm());
     84   if(rows>=2) VERIFY_IS_MUCH_SMALLER_THAN(m3.block(0,1,rows,rows-1).norm(), m3.norm());
     85   VERIFY_IS_MUCH_SMALLER_THAN(numext::imag(m3(0,0)), numext::real(m3(0,0)));
     86   VERIFY_IS_APPROX(numext::real(m3(0,0)), alpha);
     87 
     88   // test householder sequence on the left with a shift
     89 
     90   Index shift = internal::random<Index>(0, std::max<Index>(rows-2,0));
     91   Index brows = rows - shift;
     92   m1.setRandom(rows, cols);
     93   HBlockMatrixType hbm = m1.block(shift,0,brows,cols);
     94   HouseholderQR<HBlockMatrixType> qr(hbm);
     95   m2 = m1;
     96   m2.block(shift,0,brows,cols) = qr.matrixQR();
     97   HCoeffsVectorType hc = qr.hCoeffs().conjugate();
     98   HouseholderSequence<MatrixType, HCoeffsVectorType> hseq(m2, hc);
     99   hseq.setLength(hc.size()).setShift(shift);
    100   VERIFY(hseq.length() == hc.size());
    101   VERIFY(hseq.shift() == shift);
    102   
    103   MatrixType m5 = m2;
    104   m5.block(shift,0,brows,cols).template triangularView<StrictlyLower>().setZero();
    105   VERIFY_IS_APPROX(hseq * m5, m1); // test applying hseq directly
    106   m3 = hseq;
    107   VERIFY_IS_APPROX(m3 * m5, m1); // test evaluating hseq to a dense matrix, then applying
    108   
    109   SquareMatrixType hseq_mat = hseq;
    110   SquareMatrixType hseq_mat_conj = hseq.conjugate();
    111   SquareMatrixType hseq_mat_adj = hseq.adjoint();
    112   SquareMatrixType hseq_mat_trans = hseq.transpose();
    113   SquareMatrixType m6 = SquareMatrixType::Random(rows, rows);
    114   VERIFY_IS_APPROX(hseq_mat.adjoint(),    hseq_mat_adj);
    115   VERIFY_IS_APPROX(hseq_mat.conjugate(),  hseq_mat_conj);
    116   VERIFY_IS_APPROX(hseq_mat.transpose(),  hseq_mat_trans);
    117   VERIFY_IS_APPROX(hseq * m6,             hseq_mat * m6);
    118   VERIFY_IS_APPROX(hseq.adjoint() * m6,   hseq_mat_adj * m6);
    119   VERIFY_IS_APPROX(hseq.conjugate() * m6, hseq_mat_conj * m6);
    120   VERIFY_IS_APPROX(hseq.transpose() * m6, hseq_mat_trans * m6);
    121   VERIFY_IS_APPROX(m6 * hseq,             m6 * hseq_mat);
    122   VERIFY_IS_APPROX(m6 * hseq.adjoint(),   m6 * hseq_mat_adj);
    123   VERIFY_IS_APPROX(m6 * hseq.conjugate(), m6 * hseq_mat_conj);
    124   VERIFY_IS_APPROX(m6 * hseq.transpose(), m6 * hseq_mat_trans);
    125 
    126   // test householder sequence on the right with a shift
    127 
    128   TMatrixType tm2 = m2.transpose();
    129   HouseholderSequence<TMatrixType, HCoeffsVectorType, OnTheRight> rhseq(tm2, hc);
    130   rhseq.setLength(hc.size()).setShift(shift);
    131   VERIFY_IS_APPROX(rhseq * m5, m1); // test applying rhseq directly
    132   m3 = rhseq;
    133   VERIFY_IS_APPROX(m3 * m5, m1); // test evaluating rhseq to a dense matrix, then applying
    134 }
    135 
    136 EIGEN_DECLARE_TEST(householder)
    137 {
    138   for(int i = 0; i < g_repeat; i++) {
    139     CALL_SUBTEST_1( householder(Matrix<double,2,2>()) );
    140     CALL_SUBTEST_2( householder(Matrix<float,2,3>()) );
    141     CALL_SUBTEST_3( householder(Matrix<double,3,5>()) );
    142     CALL_SUBTEST_4( householder(Matrix<float,4,4>()) );
    143     CALL_SUBTEST_5( householder(MatrixXd(internal::random<int>(1,EIGEN_TEST_MAX_SIZE),internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
    144     CALL_SUBTEST_6( householder(MatrixXcf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE),internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
    145     CALL_SUBTEST_7( householder(MatrixXf(internal::random<int>(1,EIGEN_TEST_MAX_SIZE),internal::random<int>(1,EIGEN_TEST_MAX_SIZE))) );
    146     CALL_SUBTEST_8( householder(Matrix<double,1,1>()) );
    147   }
    148 }