eigen/Eigen/src/SVD/SVD.h
2009-08-16 00:02:36 +02:00

536 lines
15 KiB
C++

// This file is part of Eigen, a lightweight C++ template library
// for linear algebra.
//
// Copyright (C) 2008-2009 Gael Guennebaud <g.gael@free.fr>
//
// Eigen is free software; you can redistribute it and/or
// modify it under the terms of the GNU Lesser General Public
// License as published by the Free Software Foundation; either
// version 3 of the License, or (at your option) any later version.
//
// Alternatively, you can redistribute it and/or
// modify it under the terms of the GNU General Public License as
// published by the Free Software Foundation; either version 2 of
// the License, or (at your option) any later version.
//
// Eigen is distributed in the hope that it will be useful, but WITHOUT ANY
// WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
// FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License or the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU Lesser General Public
// License and a copy of the GNU General Public License along with
// Eigen. If not, see <http://www.gnu.org/licenses/>.
#ifndef EIGEN_SVD_H
#define EIGEN_SVD_H
/** \ingroup SVD_Module
* \nonstableyet
*
* \class SVD
*
* \brief Standard SVD decomposition of a matrix and associated features
*
* \param MatrixType the type of the matrix of which we are computing the SVD decomposition
*
* This class performs a standard SVD decomposition of a real matrix A of size \c M x \c N.
*
* \sa MatrixBase::SVD()
*/
template<typename MatrixType> class SVD
{
private:
typedef typename MatrixType::Scalar Scalar;
typedef typename NumTraits<typename MatrixType::Scalar>::Real RealScalar;
enum {
PacketSize = ei_packet_traits<Scalar>::size,
AlignmentMask = int(PacketSize)-1,
MinSize = EIGEN_ENUM_MIN(MatrixType::RowsAtCompileTime, MatrixType::ColsAtCompileTime)
};
typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, 1> ColVector;
typedef Matrix<Scalar, MatrixType::ColsAtCompileTime, 1> RowVector;
typedef Matrix<Scalar, MatrixType::RowsAtCompileTime, MatrixType::RowsAtCompileTime> MatrixUType;
typedef Matrix<Scalar, MatrixType::ColsAtCompileTime, MatrixType::ColsAtCompileTime> MatrixVType;
typedef Matrix<Scalar, MatrixType::ColsAtCompileTime, 1> SingularValuesType;
public:
/**
* \brief Default Constructor.
*
* The default constructor is useful in cases in which the user intends to
* perform decompositions via SVD::compute(const MatrixType&).
*/
SVD() : m_matU(), m_matV(), m_sigma(), m_isInitialized(false) {}
SVD(const MatrixType& matrix)
: m_matU(matrix.rows(), matrix.rows()),
m_matV(matrix.cols(),matrix.cols()),
m_sigma(matrix.cols()),
m_isInitialized(false)
{
compute(matrix);
}
template<typename OtherDerived, typename ResultType>
bool solve(const MatrixBase<OtherDerived> &b, ResultType* result) const;
const MatrixUType& matrixU() const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
return m_matU;
}
const SingularValuesType& singularValues() const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
return m_sigma;
}
const MatrixVType& matrixV() const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
return m_matV;
}
void compute(const MatrixType& matrix);
template<typename UnitaryType, typename PositiveType>
void computeUnitaryPositive(UnitaryType *unitary, PositiveType *positive) const;
template<typename PositiveType, typename UnitaryType>
void computePositiveUnitary(PositiveType *positive, UnitaryType *unitary) const;
template<typename RotationType, typename ScalingType>
void computeRotationScaling(RotationType *unitary, ScalingType *positive) const;
template<typename ScalingType, typename RotationType>
void computeScalingRotation(ScalingType *positive, RotationType *unitary) const;
protected:
// Computes (a^2 + b^2)^(1/2) without destructive underflow or overflow.
inline static Scalar pythag(Scalar a, Scalar b)
{
Scalar abs_a = ei_abs(a);
Scalar abs_b = ei_abs(b);
if (abs_a > abs_b)
return abs_a*ei_sqrt(Scalar(1.0)+ei_abs2(abs_b/abs_a));
else
return (abs_b == Scalar(0.0) ? Scalar(0.0) : abs_b*ei_sqrt(Scalar(1.0)+ei_abs2(abs_a/abs_b)));
}
inline static Scalar sign(Scalar a, Scalar b)
{
return (b >= Scalar(0.0) ? ei_abs(a) : -ei_abs(a));
}
protected:
/** \internal */
MatrixUType m_matU;
/** \internal */
MatrixVType m_matV;
/** \internal */
SingularValuesType m_sigma;
bool m_isInitialized;
};
/** Computes / recomputes the SVD decomposition A = U S V^* of \a matrix
*
* \note this code has been adapted from Numerical Recipes, third edition.
*/
template<typename MatrixType>
void SVD<MatrixType>::compute(const MatrixType& matrix)
{
const int m = matrix.rows();
const int n = matrix.cols();
m_matU.resize(m, m);
m_matU.setZero();
m_sigma.resize(n);
m_matV.resize(n,n);
int max_iters = 30;
MatrixVType& V = m_matV;
MatrixType A = matrix;
SingularValuesType& W = m_sigma;
bool flag;
int i,its,j,jj,k,l,nm;
Scalar anorm, c, f, g, h, s, scale, x, y, z;
bool convergence = true;
Scalar eps = precision<Scalar>();
Matrix<Scalar,Dynamic,1> rv1(n);
g = scale = anorm = 0;
// Householder reduction to bidiagonal form.
for (i=0; i<n; i++)
{
l = i+2;
rv1[i] = scale*g;
g = s = scale = 0.0;
if (i < m)
{
scale = A.col(i).end(m-i).cwise().abs().sum();
if (scale != Scalar(0))
{
for (k=i; k<m; k++)
{
A(k, i) /= scale;
s += A(k, i)*A(k, i);
}
f = A(i, i);
g = -sign( ei_sqrt(s), f );
h = f*g - s;
A(i, i)=f-g;
for (j=l-1; j<n; j++)
{
s = A.col(j).end(m-i).dot(A.col(i).end(m-i));
f = s/h;
A.col(j).end(m-i) += f*A.col(i).end(m-i);
}
A.col(i).end(m-i) *= scale;
}
}
W[i] = scale * g;
g = s = scale = 0.0;
if (i+1 <= m && i+1 != n)
{
scale = A.row(i).end(n-l+1).cwise().abs().sum();
if (scale != Scalar(0))
{
for (k=l-1; k<n; k++)
{
A(i, k) /= scale;
s += A(i, k)*A(i, k);
}
f = A(i,l-1);
g = -sign(ei_sqrt(s),f);
h = f*g - s;
A(i,l-1) = f-g;
rv1.end(n-l+1) = A.row(i).end(n-l+1)/h;
for (j=l-1; j<m; j++)
{
s = A.row(i).end(n-l+1).dot(A.row(j).end(n-l+1));
A.row(j).end(n-l+1) += s*rv1.end(n-l+1).transpose();
}
A.row(i).end(n-l+1) *= scale;
}
}
anorm = std::max( anorm, (ei_abs(W[i])+ei_abs(rv1[i])) );
}
// Accumulation of right-hand transformations.
for (i=n-1; i>=0; i--)
{
//Accumulation of right-hand transformations.
if (i < n-1)
{
if (g != Scalar(0.0))
{
for (j=l; j<n; j++) //Double division to avoid possible underflow.
V(j, i) = (A(i, j)/A(i, l))/g;
for (j=l; j<n; j++)
{
s = V.col(j).end(n-l).dot(A.row(i).end(n-l));
V.col(j).end(n-l) += s * V.col(i).end(n-l);
}
}
V.row(i).end(n-l).setZero();
V.col(i).end(n-l).setZero();
}
V(i, i) = 1.0;
g = rv1[i];
l = i;
}
// Accumulation of left-hand transformations.
for (i=std::min(m,n)-1; i>=0; i--)
{
l = i+1;
g = W[i];
if (n-l>0)
A.row(i).end(n-l).setZero();
if (g != Scalar(0.0))
{
g = Scalar(1.0)/g;
if (m-l)
{
for (j=l; j<n; j++)
{
s = A.col(j).end(m-l).dot(A.col(i).end(m-l));
f = (s/A(i,i))*g;
A.col(j).end(m-i) += f * A.col(i).end(m-i);
}
}
A.col(i).end(m-i) *= g;
}
else
A.col(i).end(m-i).setZero();
++A(i,i);
}
// Diagonalization of the bidiagonal form: Loop over
// singular values, and over allowed iterations.
for (k=n-1; k>=0; k--)
{
for (its=0; its<max_iters; its++)
{
flag = true;
for (l=k; l>=0; l--)
{
// Test for splitting.
nm = l-1;
// Note that rv1[1] is always zero.
//if ((double)(ei_abs(rv1[l])+anorm) == anorm)
if (l==0 || ei_abs(rv1[l]) <= eps*anorm)
{
flag = false;
break;
}
//if ((double)(ei_abs(W[nm])+anorm) == anorm)
if (ei_abs(W[nm]) <= eps*anorm)
break;
}
if (flag)
{
c = 0.0; //Cancellation of rv1[l], if l > 0.
s = 1.0;
for (i=l ;i<k+1; i++)
{
f = s*rv1[i];
rv1[i] = c*rv1[i];
//if ((double)(ei_abs(f)+anorm) == anorm)
if (ei_abs(f) <= eps*anorm)
break;
g = W[i];
h = pythag(f,g);
W[i] = h;
h = Scalar(1.0)/h;
c = g*h;
s = -f*h;
V.applyJacobiOnTheRight(i,nm,c,s);
}
}
z = W[k];
if (l == k) //Convergence.
{
if (z < 0.0) { // Singular value is made nonnegative.
W[k] = -z;
V.col(k) = -V.col(k);
}
break;
}
if (its+1 == max_iters)
{
convergence = false;
}
x = W[l]; // Shift from bottom 2-by-2 minor.
nm = k-1;
y = W[nm];
g = rv1[nm];
h = rv1[k];
f = ((y-z)*(y+z) + (g-h)*(g+h))/(Scalar(2.0)*h*y);
g = pythag(f,1.0);
f = ((x-z)*(x+z) + h*((y/(f+sign(g,f)))-h))/x;
c = s = 1.0;
//Next QR transformation:
for (j=l; j<=nm; j++)
{
i = j+1;
g = rv1[i];
y = W[i];
h = s*g;
g = c*g;
z = pythag(f,h);
rv1[j] = z;
c = f/z;
s = h/z;
f = x*c + g*s;
g = g*c - x*s;
h = y*s;
y *= c;
V.applyJacobiOnTheRight(i,j,c,s);
z = pythag(f,h);
W[j] = z;
// Rotation can be arbitrary if z = 0.
if (z!=Scalar(0))
{
z = Scalar(1.0)/z;
c = f*z;
s = h*z;
}
f = c*g + s*y;
x = c*y - s*g;
A.applyJacobiOnTheRight(i,j,c,s);
}
rv1[l] = 0.0;
rv1[k] = f;
W[k] = x;
}
}
// sort the singular values:
{
for (int i=0; i<n; i++)
{
int k;
W.end(n-i).maxCoeff(&k);
if (k != 0)
{
k += i;
std::swap(W[k],W[i]);
A.col(i).swap(A.col(k));
V.col(i).swap(V.col(k));
}
}
}
m_matU.setZero();
if (m>=n)
m_matU.block(0,0,m,n) = A;
else
m_matU = A.block(0,0,m,m);
m_isInitialized = true;
}
/** \returns the solution of \f$ A x = b \f$ using the current SVD decomposition of A.
* The parts of the solution corresponding to zero singular values are ignored.
*
* \sa MatrixBase::svd(), LU::solve(), LLT::solve()
*/
template<typename MatrixType>
template<typename OtherDerived, typename ResultType>
bool SVD<MatrixType>::solve(const MatrixBase<OtherDerived> &b, ResultType* result) const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
const int rows = m_matU.rows();
ei_assert(b.rows() == rows);
result->resize(m_matV.rows(), b.cols());
Scalar maxVal = m_sigma.cwise().abs().maxCoeff();
for (int j=0; j<b.cols(); ++j)
{
Matrix<Scalar,MatrixUType::RowsAtCompileTime,1> aux = m_matU.transpose() * b.col(j);
for (int i = 0; i <m_matU.cols(); ++i)
{
Scalar si = m_sigma.coeff(i);
if (ei_isMuchSmallerThan(ei_abs(si),maxVal))
aux.coeffRef(i) = 0;
else
aux.coeffRef(i) /= si;
}
result->col(j) = m_matV * aux;
}
return true;
}
/** Computes the polar decomposition of the matrix, as a product unitary x positive.
*
* If either pointer is zero, the corresponding computation is skipped.
*
* Only for square matrices.
*
* \sa computePositiveUnitary(), computeRotationScaling()
*/
template<typename MatrixType>
template<typename UnitaryType, typename PositiveType>
void SVD<MatrixType>::computeUnitaryPositive(UnitaryType *unitary,
PositiveType *positive) const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
ei_assert(m_matU.cols() == m_matV.cols() && "Polar decomposition is only for square matrices");
if(unitary) *unitary = m_matU * m_matV.adjoint();
if(positive) *positive = m_matV * m_sigma.asDiagonal() * m_matV.adjoint();
}
/** Computes the polar decomposition of the matrix, as a product positive x unitary.
*
* If either pointer is zero, the corresponding computation is skipped.
*
* Only for square matrices.
*
* \sa computeUnitaryPositive(), computeRotationScaling()
*/
template<typename MatrixType>
template<typename UnitaryType, typename PositiveType>
void SVD<MatrixType>::computePositiveUnitary(UnitaryType *positive,
PositiveType *unitary) const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
ei_assert(m_matU.rows() == m_matV.rows() && "Polar decomposition is only for square matrices");
if(unitary) *unitary = m_matU * m_matV.adjoint();
if(positive) *positive = m_matU * m_sigma.asDiagonal() * m_matU.adjoint();
}
/** decomposes the matrix as a product rotation x scaling, the scaling being
* not necessarily positive.
*
* If either pointer is zero, the corresponding computation is skipped.
*
* This method requires the Geometry module.
*
* \sa computeScalingRotation(), computeUnitaryPositive()
*/
template<typename MatrixType>
template<typename RotationType, typename ScalingType>
void SVD<MatrixType>::computeRotationScaling(RotationType *rotation, ScalingType *scaling) const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
ei_assert(m_matU.rows() == m_matV.rows() && "Polar decomposition is only for square matrices");
Scalar x = (m_matU * m_matV.adjoint()).determinant(); // so x has absolute value 1
Matrix<Scalar, MatrixType::RowsAtCompileTime, 1> sv(m_sigma);
sv.coeffRef(0) *= x;
if(scaling) scaling->lazyAssign(m_matV * sv.asDiagonal() * m_matV.adjoint());
if(rotation)
{
MatrixType m(m_matU);
m.col(0) /= x;
rotation->lazyAssign(m * m_matV.adjoint());
}
}
/** decomposes the matrix as a product scaling x rotation, the scaling being
* not necessarily positive.
*
* If either pointer is zero, the corresponding computation is skipped.
*
* This method requires the Geometry module.
*
* \sa computeRotationScaling(), computeUnitaryPositive()
*/
template<typename MatrixType>
template<typename ScalingType, typename RotationType>
void SVD<MatrixType>::computeScalingRotation(ScalingType *scaling, RotationType *rotation) const
{
ei_assert(m_isInitialized && "SVD is not initialized.");
ei_assert(m_matU.rows() == m_matV.rows() && "Polar decomposition is only for square matrices");
Scalar x = (m_matU * m_matV.adjoint()).determinant(); // so x has absolute value 1
Matrix<Scalar, MatrixType::RowsAtCompileTime, 1> sv(m_sigma);
sv.coeffRef(0) *= x;
if(scaling) scaling->lazyAssign(m_matU * sv.asDiagonal() * m_matU.adjoint());
if(rotation)
{
MatrixType m(m_matU);
m.col(0) /= x;
rotation->lazyAssign(m * m_matV.adjoint());
}
}
/** \svd_module
* \returns the SVD decomposition of \c *this
*/
template<typename Derived>
inline SVD<typename MatrixBase<Derived>::PlainMatrixType>
MatrixBase<Derived>::svd() const
{
return SVD<PlainMatrixType>(derived());
}
#endif // EIGEN_SVD_H