blob: 4c44b1aa6eb6b613204aac344c621a8a0012a400 [file] [edit]
// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008 Gael Guennebaud <gael.guennebaud@inria.fr>
// Copyright (C) 2009 Benoit Jacob <jacob.benoit.1@gmail.com>
//
// This Source Code Form is subject to the terms of the Mozilla
// Public License v. 2.0. If a copy of the MPL was not distributed
// with this file, You can obtain one at http://mozilla.org/MPL/2.0/.
// SPDX-License-Identifier: MPL-2.0
#include "main.h"
#include <Eigen/QR>
#include <Eigen/SVD>
#include "solverbase.h"
#include "fp_control.h"
template <typename MatrixType>
void cod() {
Index rows = internal::random<Index>(2, EIGEN_TEST_MAX_SIZE);
Index cols = internal::random<Index>(2, EIGEN_TEST_MAX_SIZE);
Index cols2 = internal::random<Index>(2, EIGEN_TEST_MAX_SIZE);
Index rank = internal::random<Index>(1, (std::min)(rows, cols) - 1);
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
using MatrixQType = Matrix<Scalar, MatrixType::RowsAtCompileTime, MatrixType::RowsAtCompileTime>;
MatrixType matrix;
createRandomPIMatrixOfRank(rank, rows, cols, matrix);
CompleteOrthogonalDecomposition<MatrixType> cod(matrix);
VERIFY(rank == cod.rank());
VERIFY(cols - cod.rank() == cod.dimensionOfKernel());
VERIFY(!cod.isInjective());
VERIFY(!cod.isInvertible());
VERIFY(!cod.isSurjective());
MatrixQType q = cod.householderQ();
VERIFY_IS_UNITARY(q);
MatrixType z = cod.matrixZ();
VERIFY_IS_UNITARY(z);
MatrixType t;
t.setZero(rows, cols);
t.topLeftCorner(rank, rank) = cod.matrixT().topLeftCorner(rank, rank).template triangularView<Upper>();
MatrixType c = q * t * z * cod.colsPermutation().inverse();
VERIFY_IS_APPROX(matrix, c);
check_solverbase<MatrixType, MatrixType>(matrix, cod, rows, cols, cols2);
// Verify that we get the same minimum-norm solution as the SVD.
MatrixType exact_solution = MatrixType::Random(cols, cols2);
MatrixType rhs = matrix * exact_solution;
MatrixType cod_solution = cod.solve(rhs);
JacobiSVD<MatrixType, ComputeThinU | ComputeThinV> svd(matrix);
MatrixType svd_solution = svd.solve(rhs);
VERIFY_IS_APPROX(cod_solution, svd_solution);
MatrixType pinv = cod.pseudoInverse();
VERIFY_IS_APPROX(cod_solution, pinv * rhs);
// Z is the identity at full column rank, also when reusing a rank-deficient decomposition of the same shape.
MatrixType deficient;
createRandomPIMatrixOfRank(cols - 1, cols, cols, deficient);
CompleteOrthogonalDecomposition<MatrixType> cod3(deficient);
cod3.compute(MatrixType::Identity(cols, cols));
VERIFY_IS_APPROX(MatrixType(cod3.matrixZ()), MatrixType::Identity(cols, cols));
// now construct a (square) matrix with prescribed determinant
Index size = internal::random<Index>(2, 20);
matrix.setZero(size, size);
setRandomWellConditionedDiagonal(matrix);
Scalar det = matrix.diagonal().prod();
RealScalar absdet = numext::abs(det);
// Rotate the diagonal matrix by a random unitary matrix, so that the determinants are those of a dense matrix.
q = generateRandomUnitaryMatrix<MatrixQType>(size);
matrix = q * matrix * q.adjoint();
CompleteOrthogonalDecomposition<MatrixType> cod2(matrix);
VERIFY_IS_APPROX(det, cod2.determinant());
VERIFY_IS_APPROX(absdet, cod2.absDeterminant());
VERIFY_IS_APPROX(numext::log(absdet), cod2.logAbsDeterminant());
VERIFY_IS_APPROX(numext::sign(det), cod2.signDeterminant());
}
template <typename MatrixType, int Cols2>
void cod_fixedsize() {
enum { Rows = MatrixType::RowsAtCompileTime, Cols = MatrixType::ColsAtCompileTime };
using Scalar = typename MatrixType::Scalar;
using COD = CompleteOrthogonalDecomposition<Matrix<Scalar, Rows, Cols>>;
int rank = internal::random<int>(1, (std::min)(int(Rows), int(Cols)) - 1);
Matrix<Scalar, Rows, Cols> matrix;
createRandomPIMatrixOfRank(rank, Rows, Cols, matrix);
COD cod(matrix);
VERIFY(rank == cod.rank());
VERIFY(Cols - cod.rank() == cod.dimensionOfKernel());
VERIFY(cod.isInjective() == (rank == Rows));
VERIFY(cod.isSurjective() == (rank == Cols));
VERIFY(cod.isInvertible() == (cod.isInjective() && cod.isSurjective()));
Matrix<Scalar, Cols, Cols> z = cod.matrixZ();
VERIFY_IS_UNITARY(z);
check_solverbase<Matrix<Scalar, Cols, Cols2>, Matrix<Scalar, Rows, Cols2> >(matrix, cod, Rows, Cols, Cols2);
// Verify that we get the same minimum-norm solution as the SVD.
Matrix<Scalar, Cols, Cols2> exact_solution;
exact_solution.setRandom(Cols, Cols2);
Matrix<Scalar, Rows, Cols2> rhs = matrix * exact_solution;
Matrix<Scalar, Cols, Cols2> cod_solution = cod.solve(rhs);
JacobiSVD<MatrixType, ComputeFullU | ComputeFullV> svd(matrix);
Matrix<Scalar, Cols, Cols2> svd_solution = svd.solve(rhs);
VERIFY_IS_APPROX(cod_solution, svd_solution);
typename Inverse<COD>::PlainObject pinv = cod.pseudoInverse();
VERIFY_IS_APPROX(cod_solution, pinv * rhs);
}
template <typename MatrixType>
void qr() {
using std::sqrt;
Index rows = internal::random<Index>(2, EIGEN_TEST_MAX_SIZE), cols = internal::random<Index>(2, EIGEN_TEST_MAX_SIZE),
cols2 = internal::random<Index>(2, EIGEN_TEST_MAX_SIZE);
Index rank = internal::random<Index>(1, (std::min)(rows, cols) - 1);
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
using MatrixQType = Matrix<Scalar, MatrixType::RowsAtCompileTime, MatrixType::RowsAtCompileTime>;
MatrixType m1;
createRandomPIMatrixOfRank(rank, rows, cols, m1);
ColPivHouseholderQR<MatrixType> qr(m1);
VERIFY_IS_EQUAL(rank, qr.rank());
VERIFY_IS_EQUAL(cols - qr.rank(), qr.dimensionOfKernel());
VERIFY(!qr.isInjective());
VERIFY(!qr.isInvertible());
VERIFY(!qr.isSurjective());
MatrixQType q = qr.householderQ();
VERIFY_IS_UNITARY(q);
MatrixType r = qr.matrixQR().template triangularView<Upper>();
MatrixType c = q * r * qr.colsPermutation().inverse();
VERIFY_IS_APPROX(m1, c);
// Verify that the absolute value of the diagonal elements in R are
// non-increasing until they reach the singularity threshold.
RealScalar threshold = sqrt(RealScalar(rows)) * numext::abs(r(0, 0)) * NumTraits<Scalar>::epsilon();
for (Index i = 0; i < (std::min)(rows, cols) - 1; ++i) {
RealScalar x = numext::abs(r(i, i));
RealScalar y = numext::abs(r(i + 1, i + 1));
if (x < threshold && y < threshold) continue;
if (!test_isApproxOrLessThan(y, x)) {
for (Index j = 0; j < (std::min)(rows, cols); ++j) {
std::cout << "i = " << j << ", |r_ii| = " << numext::abs(r(j, j)) << std::endl;
}
std::cout << "Failure at i=" << i << ", rank=" << rank << ", threshold=" << threshold << std::endl;
}
VERIFY_IS_APPROX_OR_LESS_THAN(y, x);
}
check_solverbase<MatrixType, MatrixType>(m1, qr, rows, cols, cols2);
{
MatrixType m2, m3;
Index size = rows;
// Create a random diagonally dominant (thus invertible) matrix.
m1 = MatrixType::Random(size, size);
m1.diagonal().array() += Scalar(2 * size);
qr.compute(m1);
MatrixType m1_inv = qr.inverse();
m3 = m1 * MatrixType::Random(size, cols2);
m2 = qr.solve(m3);
VERIFY_IS_APPROX(m2, m1_inv * m3);
}
}
template <typename MatrixType, int Cols2>
void qr_fixedsize() {
using std::abs;
using std::sqrt;
enum { Rows = MatrixType::RowsAtCompileTime, Cols = MatrixType::ColsAtCompileTime };
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
int rank = internal::random<int>(1, (std::min)(int(Rows), int(Cols)) - 1);
Matrix<Scalar, Rows, Cols> m1;
createRandomPIMatrixOfRank(rank, Rows, Cols, m1);
ColPivHouseholderQR<Matrix<Scalar, Rows, Cols> > qr(m1);
VERIFY_IS_EQUAL(rank, qr.rank());
VERIFY_IS_EQUAL(Cols - qr.rank(), qr.dimensionOfKernel());
VERIFY_IS_EQUAL(qr.isInjective(), (rank == Rows));
VERIFY_IS_EQUAL(qr.isSurjective(), (rank == Cols));
VERIFY_IS_EQUAL(qr.isInvertible(), (qr.isInjective() && qr.isSurjective()));
Matrix<Scalar, Rows, Cols> r = qr.matrixQR().template triangularView<Upper>();
Matrix<Scalar, Rows, Cols> c = qr.householderQ() * r * qr.colsPermutation().inverse();
VERIFY_IS_APPROX(m1, c);
check_solverbase<Matrix<Scalar, Cols, Cols2>, Matrix<Scalar, Rows, Cols2> >(m1, qr, Rows, Cols, Cols2);
// Verify that the absolute value of the diagonal elements in R are
// non-increasing until they reache the singularity threshold.
RealScalar threshold = sqrt(RealScalar(Rows)) * (std::abs)(r(0, 0)) * NumTraits<Scalar>::epsilon();
for (Index i = 0; i < (std::min)(int(Rows), int(Cols)) - 1; ++i) {
RealScalar x = numext::abs(r(i, i));
RealScalar y = numext::abs(r(i + 1, i + 1));
if (x < threshold && y < threshold) continue;
if (!test_isApproxOrLessThan(y, x)) {
for (Index j = 0; j < (std::min)(int(Rows), int(Cols)); ++j) {
std::cout << "i = " << j << ", |r_ii| = " << numext::abs(r(j, j)) << std::endl;
}
std::cout << "Failure at i=" << i << ", rank=" << rank << ", threshold=" << threshold << std::endl;
}
VERIFY_IS_APPROX_OR_LESS_THAN(y, x);
}
}
// This test is meant to verify that pivots are chosen such that
// even for a graded matrix, the diagonal of R falls of roughly
// monotonically until it reaches the threshold for singularity.
// We use the so-called Kahan matrix, which is a famous counter-example
// for rank-revealing QR. See
// http://www.netlib.org/lapack/lawnspdf/lawn176.pdf
// page 3 for more detail.
template <typename MatrixType>
void qr_kahan_matrix() {
using std::abs;
using std::sqrt;
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
Index rows = 300, cols = rows;
MatrixType m1;
m1.setZero(rows, cols);
RealScalar s = std::pow(NumTraits<RealScalar>::epsilon(), 1.0 / rows);
RealScalar c = std::sqrt(1 - s * s);
RealScalar pow_s_i(1.0); // pow(s,i)
for (Index i = 0; i < rows; ++i) {
m1(i, i) = pow_s_i;
m1.row(i).tail(rows - i - 1) = -pow_s_i * c * MatrixType::Ones(1, rows - i - 1);
pow_s_i *= s;
}
m1 = (m1 + m1.transpose()).eval();
ColPivHouseholderQR<MatrixType> qr(m1);
MatrixType r = qr.matrixQR().template triangularView<Upper>();
RealScalar threshold = std::sqrt(RealScalar(rows)) * numext::abs(r(0, 0)) * NumTraits<Scalar>::epsilon();
for (Index i = 0; i < (std::min)(rows, cols) - 1; ++i) {
RealScalar x = numext::abs(r(i, i));
RealScalar y = numext::abs(r(i + 1, i + 1));
if (x < threshold && y < threshold) continue;
if (!test_isApproxOrLessThan(y, x)) {
for (Index j = 0; j < (std::min)(rows, cols); ++j) {
std::cout << "i = " << j << ", |r_ii| = " << numext::abs(r(j, j)) << std::endl;
}
std::cout << "Failure at i=" << i << ", rank=" << qr.rank() << ", threshold=" << threshold << std::endl;
}
VERIFY_IS_APPROX_OR_LESS_THAN(y, x);
}
}
template <typename MatrixType>
void qr_invertible() {
using std::abs;
using std::log;
using RealScalar = typename NumTraits<typename MatrixType::Scalar>::Real;
using Scalar = typename MatrixType::Scalar;
int size = internal::random<int>(10, 50);
MatrixType m1(size, size), m2(size, size), m3(size, size);
m1 = MatrixType::Random(size, size);
if (std::is_same<RealScalar, float>::value) {
// let's build a matrix more stable to inverse
MatrixType a = MatrixType::Random(size, size * 2);
m1 += a * a.adjoint();
}
ColPivHouseholderQR<MatrixType> qr(m1);
check_solverbase<MatrixType, MatrixType>(m1, qr, size, size, size);
// now construct a matrix with prescribed determinant
m1.setZero();
setRandomWellConditionedDiagonal(m1);
Scalar det = m1.diagonal().prod();
RealScalar absdet = abs(det);
m3 = qr.householderQ(); // get a unitary
m1 = m3 * m1 * m3.adjoint();
qr.compute(m1);
VERIFY_IS_APPROX(det, qr.determinant());
VERIFY_IS_APPROX(absdet, qr.absDeterminant());
VERIFY_IS_APPROX(log(absdet), qr.logAbsDeterminant());
VERIFY_IS_APPROX(numext::sign(det), qr.signDeterminant());
}
template <typename MatrixType>
void qr_verify_assert() {
MatrixType tmp;
ColPivHouseholderQR<MatrixType> qr;
VERIFY_RAISES_ASSERT(qr.matrixQR())
VERIFY_RAISES_ASSERT(qr.solve(tmp))
VERIFY_RAISES_ASSERT(qr.transpose().solve(tmp))
VERIFY_RAISES_ASSERT(qr.adjoint().solve(tmp))
VERIFY_RAISES_ASSERT(qr.householderQ())
VERIFY_RAISES_ASSERT(qr.dimensionOfKernel())
VERIFY_RAISES_ASSERT(qr.isInjective())
VERIFY_RAISES_ASSERT(qr.isSurjective())
VERIFY_RAISES_ASSERT(qr.isInvertible())
VERIFY_RAISES_ASSERT(qr.inverse())
VERIFY_RAISES_ASSERT(qr.determinant())
VERIFY_RAISES_ASSERT(qr.absDeterminant())
VERIFY_RAISES_ASSERT(qr.logAbsDeterminant())
VERIFY_RAISES_ASSERT(qr.signDeterminant())
}
template <typename MatrixType>
void cod_verify_assert() {
MatrixType tmp;
CompleteOrthogonalDecomposition<MatrixType> cod;
VERIFY_RAISES_ASSERT(cod.matrixQTZ())
VERIFY_RAISES_ASSERT(cod.solve(tmp))
VERIFY_RAISES_ASSERT(cod.transpose().solve(tmp))
VERIFY_RAISES_ASSERT(cod.adjoint().solve(tmp))
VERIFY_RAISES_ASSERT(cod.householderQ())
VERIFY_RAISES_ASSERT(cod.dimensionOfKernel())
VERIFY_RAISES_ASSERT(cod.isInjective())
VERIFY_RAISES_ASSERT(cod.isSurjective())
VERIFY_RAISES_ASSERT(cod.isInvertible())
VERIFY_RAISES_ASSERT(cod.pseudoInverse())
VERIFY_RAISES_ASSERT(cod.determinant())
VERIFY_RAISES_ASSERT(cod.absDeterminant())
VERIFY_RAISES_ASSERT(cod.logAbsDeterminant())
VERIFY_RAISES_ASSERT(cod.signDeterminant())
}
// Stress test: verify rank detection on partial isometries (SVs = 0 or 1) across
// many random trials and various aspect ratios (square, tall, wide).
// This tests ROBUSTNESS: the threshold must be large enough that roundoff noise
// in the null-space R diagonal elements does not cause rank overestimation.
template <typename MatrixType>
void qr_rank_detection_stress() {
// Test a range of matrix sizes and aspect ratios.
const Index sizes[][2] = {{10, 10}, {20, 20}, {50, 50}, {100, 100}, {40, 10}, {100, 10}, {10, 40}, {10, 100}};
for (const auto& sz : sizes) {
const Index rows = sz[0], cols = sz[1];
const Index min_dim = (std::min)(rows, cols);
// Test several rank values: 1, half, and min_dim - 1.
for (Index rank : {Index(1), (std::max)(Index(1), min_dim / 2), min_dim - 1}) {
if (rank >= min_dim) continue;
for (int trial = 0; trial < 20; ++trial) {
MatrixType m1;
createRandomPIMatrixOfRank(rank, rows, cols, m1);
ColPivHouseholderQR<MatrixType> qr(m1);
VERIFY_IS_EQUAL(rank, qr.rank());
}
}
}
}
// Efficiency test: verify the threshold is not so large that it causes false
// rank deficiency. Creates matrices with smallest singular value well above
// the backward error bound, and verifies they are detected as full rank.
template <typename MatrixType>
void qr_threshold_efficiency() {
using RealScalar = typename MatrixType::RealScalar;
using RealVectorType = Matrix<RealScalar, Dynamic, 1>;
const Index sizes[][2] = {{10, 10}, {50, 50}, {100, 100}, {40, 10}, {10, 40}};
for (const auto& sz : sizes) {
const Index rows = sz[0], cols = sz[1];
const Index min_dim = (std::min)(rows, cols);
// Create a matrix with prescribed singular values: smallest SV = 100 * threshold.
// The threshold is 4*min(m,n)*eps, so we set sigma_min = 400*min(m,n)*eps.
// This must be detected as full rank.
RealScalar sigma_min = RealScalar(400) * RealScalar(min_dim) * NumTraits<RealScalar>::epsilon();
RealVectorType svs = setupRangeSvs<RealVectorType>(min_dim, sigma_min, RealScalar(1));
MatrixType m1;
generateRandomMatrixSvs(svs, rows, cols, m1);
ColPivHouseholderQR<MatrixType> qr(m1);
// sigma_min is 100x the threshold — must be detected as full rank.
VERIFY_IS_EQUAL(min_dim, qr.rank());
// Also check with FullPivHouseholderQR.
FullPivHouseholderQR<MatrixType> fpqr(m1);
VERIFY_IS_EQUAL(min_dim, fpqr.rank());
}
}
// Test rank detection on matrices with geometrically distributed singular values
// and a clear gap at the desired rank. This mimics real-world matrices better
// than partial isometries and stresses the threshold from both sides.
template <typename MatrixType>
void qr_rank_gap_test() {
using RealScalar = typename MatrixType::RealScalar;
using RealVectorType = Matrix<RealScalar, Dynamic, 1>;
const Index sizes[][2] = {{20, 20}, {50, 50}, {100, 100}, {50, 20}, {20, 50}};
for (const auto& sz : sizes) {
const Index rows = sz[0], cols = sz[1];
const Index min_dim = (std::min)(rows, cols);
const Index rank = (std::max)(Index(1), min_dim / 2);
// Singular values: [1, ..., sigma_rank] for the "signal" part,
// then [eps_level, ..., eps_level] for the "noise" part.
// The gap ratio is sigma_rank / eps_level >> 1.
RealScalar sigma_rank = RealScalar(0.1);
RealScalar eps_level = NumTraits<RealScalar>::epsilon();
RealVectorType svs(min_dim);
// Signal part: geometrically spaced from 1 to sigma_rank.
for (Index i = 0; i < rank; ++i) {
RealScalar t = (rank > 1) ? RealScalar(i) / RealScalar(rank - 1) : RealScalar(0);
svs(i) = std::pow(sigma_rank, t); // svs(0) = 1, svs(rank-1) = sigma_rank
}
// Noise part: near machine epsilon (well below any reasonable threshold).
for (Index i = rank; i < min_dim; ++i) {
svs(i) = eps_level * RealScalar(min_dim - i);
}
MatrixType m1;
generateRandomMatrixSvs(svs, rows, cols, m1);
ColPivHouseholderQR<MatrixType> qr(m1);
VERIFY_IS_EQUAL(rank, qr.rank());
FullPivHouseholderQR<MatrixType> fpqr(m1);
VERIFY_IS_EQUAL(rank, fpqr.rank());
}
}
// Regression: a matrix without columns has no column norm to take the maximum of.
template <typename MatrixType>
void qr_empty() {
ColPivHouseholderQR<MatrixType> qr{MatrixType(3, 0)};
VERIFY_IS_EQUAL(qr.rank(), Index(0));
qr.compute(MatrixType(0, 0));
VERIFY_IS_EQUAL(qr.rank(), Index(0));
CompleteOrthogonalDecomposition<MatrixType> cod{MatrixType(3, 0)};
VERIFY_IS_EQUAL(cod.rank(), Index(0));
}
// Column norms and the rank threshold must not square the entries: pivoting and rank detection have to work for entries
// whose squares overflow or underflow.
template <typename MatrixType>
void qr_configure_blocking(ColPivHouseholderQR<MatrixType>&) {}
template <typename QRType>
void qr_configure_blocking(QRType&) {}
template <typename MatrixType>
void qr_configure_blocking(RandColPivHouseholderQR<MatrixType>& qr) {
qr.setBlockSize(4).setSeed(0x5ca1e);
}
template <typename QRType>
void qr_extreme_scale() {
using MatrixType = typename QRType::MatrixType;
using Scalar = typename MatrixType::Scalar;
using VectorType = Matrix<Scalar, Dynamic, 1>;
const Scalar eps = NumTraits<Scalar>::epsilon();
const Scalar big = numext::pow((std::numeric_limits<Scalar>::max)(), Scalar(0.55));
for (const Scalar scale : {big, Scalar(1) / big}) {
// Rank one: only the largest column, the second one, can be a pivot.
VectorType u(6), v(4);
u << 1, 2, -1, Scalar(0.5), 3, -2;
v << Scalar(0.5), -2, 1, Scalar(0.25);
MatrixType A = u * v.transpose();
A *= scale;
QRType qr;
qr_configure_blocking(qr);
qr.compute(A);
VERIFY_IS_EQUAL(qr.rank(), Index(1));
VERIFY_IS_EQUAL(qr.nonzeroPivots(), Index(1));
VERIFY_IS_EQUAL(qr.colsPermutation().indices()(0), 1);
VERIFY((A * qr.solve(u) - u).stableNorm() <= Scalar(64) * eps * u.stableNorm());
// Rank 12 in a size for which the randomized QR takes its blocked path.
MatrixType B = MatrixType::Random(60, 12) * MatrixType::Random(12, 40);
B *= scale;
QRType qr2;
qr_configure_blocking(qr2);
qr2.compute(B);
VERIFY_IS_EQUAL(qr2.rank(), Index(12));
const MatrixType R = qr2.matrixR().template triangularView<Upper>();
VERIFY((qr2.householderQ() * R - B * qr2.colsPermutation()).stableNorm() <= Scalar(1024) * eps * B.stableNorm());
}
}
// Entries uniform in [-1, 1) from a linear congruential generator, so that the matrices of qr_flush_band are the same
// on every platform and for every seed of the test.
template <typename Scalar>
Scalar qr_lcg_scalar(typename NumTraits<Scalar>::Real re, typename NumTraits<Scalar>::Real, std::false_type) {
return Scalar(re);
}
template <typename Scalar>
Scalar qr_lcg_scalar(typename NumTraits<Scalar>::Real re, typename NumTraits<Scalar>::Real im, std::true_type) {
return Scalar(re, im);
}
template <typename MatrixType>
MatrixType qr_lcg_matrix(Index rows, Index cols, uint32_t& state) {
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
const auto next = [&state]() {
state = state * 1664525u + 1013904223u;
return RealScalar((state >> 8) & 0xffff) / RealScalar(32768) - RealScalar(1);
};
MatrixType m(rows, cols);
for (Index j = 0; j < cols; ++j)
for (Index i = 0; i < rows; ++i) {
const RealScalar re = next();
const RealScalar im = next();
m(i, j) = qr_lcg_scalar<Scalar>(re, im, internal::bool_constant<NumTraits<Scalar>::IsComplex>());
}
return m;
}
// Rank detection compares residual column norms with eps * maxnorm. For maxnorm a few 1e4 times sqrt(min) the squares
// of the entries of the residual of the last pivot are subnormal: the rank must not drop, whether or not subnormals are
// flushed to zero. A has rank 7 plus a perturbation of relative size 1e-4, which is above the default threshold; the
// cancellation in the first seven steps makes the factorization recompute the column norms.
template <typename QRType, bool Blocked>
void qr_flush_band() {
using MatrixType = typename QRType::MatrixType;
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
const Index rows = 40, rank = 7, cols = 8;
uint32_t state = 3u * 7919u;
// The factors are generated one after the other: the evaluation order of the operands of a product is unspecified.
const MatrixType B1 = qr_lcg_matrix<MatrixType>(rows, rank, state);
const MatrixType B2 = qr_lcg_matrix<MatrixType>(rank, cols, state);
const MatrixType C1 = qr_lcg_matrix<MatrixType>(rows, 1, state);
const MatrixType C2 = qr_lcg_matrix<MatrixType>(1, cols, state);
const MatrixType B = B1 * B2;
const MatrixType C = C1 * C2;
const MatrixType A = B + Scalar(RealScalar(1e-4)) * C;
const RealScalar maxnorm = A.colwise().norm().maxCoeff();
for (const RealScalar factor : {RealScalar(2e4), RealScalar(3e4), RealScalar(4.5e4)}) {
const MatrixType scaled = A * Scalar(factor * numext::sqrt((std::numeric_limits<RealScalar>::min)()) / maxnorm);
for (const FlushToZeroMode mode : {FlushToZeroMode::None, FlushToZeroMode::Both}) {
ScopedFlushToZero flush(mode);
QRType qr;
EIGEN_IF_CONSTEXPR (Blocked) qr_configure_blocking(qr);
qr.compute(scaled);
VERIFY_IS_EQUAL(qr.rank(), cols);
}
}
}
template <typename MatrixType>
void cod_extreme_scale() {
using Scalar = typename MatrixType::Scalar;
const Scalar eps = NumTraits<Scalar>::epsilon();
const Scalar big = numext::pow((std::numeric_limits<Scalar>::max)(), Scalar(0.55));
for (const Scalar scale : {big, Scalar(1) / big}) {
MatrixType A = MatrixType::Random(6, 2) * MatrixType::Random(2, 4);
A *= scale;
const MatrixType b = A * MatrixType::Ones(4, 1);
CompleteOrthogonalDecomposition<MatrixType> cod(A);
VERIFY_IS_EQUAL(cod.rank(), Index(2));
VERIFY((A * cod.solve(b) - b).stableNorm() <= Scalar(256) * eps * b.stableNorm());
}
}
// The rank, and with it the Z reflectors, of a complete orthogonal decomposition are fixed by compute(): a threshold
// set afterwards applies to the next compute() and leaves the decomposition and its solutions unchanged.
template <typename MatrixType>
void cod_set_threshold_after_compute() {
using Scalar = typename MatrixType::Scalar;
using RealScalar = typename MatrixType::RealScalar;
const Index rows = 8, cols = 6;
MatrixType A = MatrixType::Random(rows, cols);
A.col(cols - 1) = A.col(0) + Scalar(RealScalar(1e-6)) * MatrixType::Random(rows, 1);
const MatrixType b = MatrixType::Random(rows, 2);
CompleteOrthogonalDecomposition<MatrixType> cod(A);
VERIFY_IS_EQUAL(cod.rank(), cols);
const MatrixType x_full = cod.solve(b);
cod.setThreshold(RealScalar(1e-4));
VERIFY_IS_EQUAL(cod.rank(), cols);
VERIFY_IS_APPROX(cod.solve(b), x_full);
VERIFY(cod.matrixZ().isUnitary());
cod.compute(A);
VERIFY_IS_EQUAL(cod.rank(), cols - 1);
CompleteOrthogonalDecomposition<MatrixType> reference;
reference.setThreshold(RealScalar(1e-4));
reference.compute(A);
VERIFY_IS_EQUAL(reference.rank(), cols - 1);
const MatrixType x_deficient = cod.solve(b);
VERIFY_IS_APPROX(x_deficient, reference.solve(b));
VERIFY(cod.matrixZ().isUnitary());
// Going back to the default threshold does not change a decomposition that has already been computed either.
cod.setThreshold(Default);
VERIFY_IS_EQUAL(cod.rank(), cols - 1);
VERIFY_IS_APPROX(cod.solve(b), x_deficient);
VERIFY_IS_EQUAL(cod.dimensionOfKernel(), Index(1));
VERIFY(!cod.isInjective());
}
// A right-hand side without columns, and an empty matrix, give an empty solution of the right size; sanitizer builds
// reported a reference to the first element of an empty workspace.
template <typename Decomposition, typename MatrixType>
void qr_empty_rhs() {
const MatrixType A = MatrixType::Random(4, 4);
const Decomposition dec(A);
const MatrixType empty_rhs(4, 0);
VERIFY_IS_EQUAL(dec.solve(empty_rhs).rows(), 4);
VERIFY_IS_EQUAL(dec.solve(empty_rhs).cols(), 0);
VERIFY_IS_EQUAL(dec.transpose().solve(empty_rhs).rows(), 4);
VERIFY_IS_EQUAL(dec.transpose().solve(empty_rhs).cols(), 0);
VERIFY_IS_EQUAL(dec.adjoint().solve(empty_rhs).cols(), 0);
const MatrixType B = MatrixType::Random(4, 3);
const MatrixType X = dec.solve(B);
VERIFY_IS_EQUAL(X.cols(), 3);
VERIFY_IS_APPROX(A * X, B);
const Decomposition empty_dec{MatrixType(0, 0)};
VERIFY_IS_EQUAL(empty_dec.solve(MatrixType(0, 2)).rows(), 0);
VERIFY_IS_EQUAL(empty_dec.solve(MatrixType(0, 2)).cols(), 2);
VERIFY_IS_EQUAL(empty_dec.transpose().solve(MatrixType(0, 0)).cols(), 0);
}
EIGEN_DECLARE_TEST(qr_colpivoting) {
for (int i = 0; i < g_repeat; i++) {
CALL_SUBTEST_1(qr<MatrixXf>());
CALL_SUBTEST_2(qr<MatrixXd>());
CALL_SUBTEST_3(qr<MatrixXcd>());
CALL_SUBTEST_10((qr<Matrix<double, Dynamic, Dynamic, RowMajor>>()));
CALL_SUBTEST_11((qr<Matrix<std::complex<double>, Dynamic, Dynamic, RowMajor>>()));
CALL_SUBTEST_4((qr_fixedsize<Matrix<float, 3, 5>, 4>()));
CALL_SUBTEST_5((qr_fixedsize<Matrix<double, 6, 2>, 3>()));
CALL_SUBTEST_5((qr_fixedsize<Matrix<double, 1, 1>, 1>()));
}
for (int i = 0; i < g_repeat; i++) {
CALL_SUBTEST_1(cod<MatrixXf>());
CALL_SUBTEST_2(cod<MatrixXd>());
CALL_SUBTEST_3(cod<MatrixXcd>());
CALL_SUBTEST_4((cod_fixedsize<Matrix<float, 3, 5>, 4>()));
CALL_SUBTEST_5((cod_fixedsize<Matrix<double, 6, 2>, 3>()));
CALL_SUBTEST_5((cod_fixedsize<Matrix<double, 1, 1>, 1>()));
}
for (int i = 0; i < g_repeat; i++) {
CALL_SUBTEST_1(qr_invertible<MatrixXf>());
CALL_SUBTEST_2(qr_invertible<MatrixXd>());
CALL_SUBTEST_6(qr_invertible<MatrixXcf>());
CALL_SUBTEST_3(qr_invertible<MatrixXcd>());
}
CALL_SUBTEST_7(qr_verify_assert<Matrix3f>());
CALL_SUBTEST_8(qr_verify_assert<Matrix3d>());
CALL_SUBTEST_1(qr_verify_assert<MatrixXf>());
CALL_SUBTEST_2(qr_verify_assert<MatrixXd>());
CALL_SUBTEST_6(qr_verify_assert<MatrixXcf>());
CALL_SUBTEST_3(qr_verify_assert<MatrixXcd>());
CALL_SUBTEST_7(cod_verify_assert<Matrix3f>());
CALL_SUBTEST_8(cod_verify_assert<Matrix3d>());
CALL_SUBTEST_1(cod_verify_assert<MatrixXf>());
CALL_SUBTEST_2(cod_verify_assert<MatrixXd>());
CALL_SUBTEST_6(cod_verify_assert<MatrixXcf>());
CALL_SUBTEST_3(cod_verify_assert<MatrixXcd>());
// Test problem size constructors
CALL_SUBTEST_9(ColPivHouseholderQR<MatrixXf>(10, 20));
CALL_SUBTEST_2(qr_empty<MatrixXd>());
CALL_SUBTEST_2((qr_empty_rhs<ColPivHouseholderQR<MatrixXd>, MatrixXd>()));
CALL_SUBTEST_2((qr_empty_rhs<CompleteOrthogonalDecomposition<MatrixXd>, MatrixXd>()));
CALL_SUBTEST_3((qr_empty_rhs<ColPivHouseholderQR<MatrixXcd>, MatrixXcd>()));
CALL_SUBTEST_3((qr_empty_rhs<CompleteOrthogonalDecomposition<MatrixXcd>, MatrixXcd>()));
CALL_SUBTEST_2((qr_empty_rhs<RandColPivHouseholderQR<MatrixXd>, MatrixXd>()));
CALL_SUBTEST_1((qr_extreme_scale<ColPivHouseholderQR<MatrixXf>>()));
CALL_SUBTEST_2((qr_extreme_scale<ColPivHouseholderQR<MatrixXd>>()));
CALL_SUBTEST_1((qr_extreme_scale<RandColPivHouseholderQR<MatrixXf>>()));
CALL_SUBTEST_2((qr_extreme_scale<RandColPivHouseholderQR<MatrixXd>>()));
CALL_SUBTEST_1((qr_flush_band<ColPivHouseholderQR<MatrixXf>, false>()));
CALL_SUBTEST_2((qr_flush_band<ColPivHouseholderQR<MatrixXd>, false>()));
CALL_SUBTEST_1((qr_flush_band<RandColPivHouseholderQR<MatrixXf>, false>()));
CALL_SUBTEST_1((qr_flush_band<RandColPivHouseholderQR<MatrixXf>, true>()));
CALL_SUBTEST_2((qr_flush_band<RandColPivHouseholderQR<MatrixXd>, true>()));
CALL_SUBTEST_1((qr_flush_band<CompleteOrthogonalDecomposition<MatrixXf>, false>()));
CALL_SUBTEST_2((qr_flush_band<CompleteOrthogonalDecomposition<MatrixXd>, false>()));
CALL_SUBTEST_3((qr_flush_band<ColPivHouseholderQR<MatrixXcf>, false>()));
CALL_SUBTEST_10((qr_flush_band<ColPivHouseholderQR<Matrix<float, Dynamic, Dynamic, RowMajor>>, false>()));
CALL_SUBTEST_10((qr_flush_band<RandColPivHouseholderQR<Matrix<float, Dynamic, Dynamic, RowMajor>>, true>()));
CALL_SUBTEST_1(cod_extreme_scale<MatrixXf>());
CALL_SUBTEST_2(cod_extreme_scale<MatrixXd>());
CALL_SUBTEST_2(cod_set_threshold_after_compute<MatrixXd>());
CALL_SUBTEST_3(cod_set_threshold_after_compute<MatrixXcd>());
CALL_SUBTEST_1(qr_kahan_matrix<MatrixXf>());
CALL_SUBTEST_2(qr_kahan_matrix<MatrixXd>());
// Stress tests for rank detection threshold robustness and efficiency.
CALL_SUBTEST_1(qr_rank_detection_stress<MatrixXf>());
CALL_SUBTEST_2(qr_rank_detection_stress<MatrixXd>());
CALL_SUBTEST_1(qr_threshold_efficiency<MatrixXf>());
CALL_SUBTEST_2(qr_threshold_efficiency<MatrixXd>());
CALL_SUBTEST_1(qr_rank_gap_test<MatrixXf>());
CALL_SUBTEST_2(qr_rank_gap_test<MatrixXd>());
}