devtools/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2026-04-10 11:34:33 +08:00
Code Issues Packages Projects Releases Wiki Activity

port unsupported modules to new API

Browse Source
This commit is contained in:
Gael Guennebaud
2010-01-05 15:38:20 +01:00
parent cab85218db
commit 39209edd71
9 changed files with 189 additions and 189 deletions
Download Patch File Download Diff File Expand all files Collapse all files

72
unsupported/Eigen/src/MatrixFunctions/MatrixExponential.h
View File

@@ -31,16 +31,16 @@
/** \ingroup MatrixFunctions_Module
*
* \brief Compute the matrix exponential.
* \brief Compute the matrix exponential.
*
* \param M matrix whose exponential is to be computed.
* \param M matrix whose exponential is to be computed.
* \param result pointer to the matrix in which to store the result.
*
* The matrix exponential of \f$ M \f$ is defined by
* \f[ \exp(M) = \sum_{k=0}^\infty \frac{M^k}{k!}. \f]
* The matrix exponential can be used to solve linear ordinary
* differential equations: the solution of \f$ y' = My \f$ with the
* initial condition \f$ y(0) = y_0 \f$ is given by
* initial condition \f$ y(0) = y_0 \f$ is given by
* \f$ y(t) = \exp(M) y_0 \f$.
*
* The cost of the computation is approximately \f$ 20 n^3 \f$ for
@@ -54,17 +54,17 @@
* squaring. The degree of the Padé approximant is chosen such
* that the approximation error is less than the round-off
* error. However, errors may accumulate during the squaring phase.
*
*
* Details of the algorithm can be found in: Nicholas J. Higham, "The
* scaling and squaring method for the matrix exponential revisited,"
* <em>SIAM J. %Matrix Anal. Applic.</em>, <b>26</b>:1179&ndash;1193,
* 2005.
* 2005.
*
* Example: The following program checks that
* \f[ \exp \left[ \begin{array}{ccc}
* 0 & \frac14\pi & 0 \\
* \f[ \exp \left[ \begin{array}{ccc}
* 0 & \frac14\pi & 0 \\
* -\frac14\pi & 0 & 0 \\
* 0 & 0 & 0
* 0 & 0 & 0
* \end{array} \right] = \left[ \begin{array}{ccc}
* \frac12\sqrt2 & -\frac12\sqrt2 & 0 \\
* \frac12\sqrt2 & \frac12\sqrt2 & 0 \\
@@ -76,11 +76,11 @@
* \include MatrixExponential.cpp
* Output: \verbinclude MatrixExponential.out
*
* \note \p M has to be a matrix of \c float, \c double,
* \note \p M has to be a matrix of \c float, \c double,
* \c complex<float> or \c complex<double> .
*/
template <typename Derived>
EIGEN_STRONG_INLINE void ei_matrix_exponential(const MatrixBase<Derived> &M,
EIGEN_STRONG_INLINE void ei_matrix_exponential(const MatrixBase<Derived> &M,
typename MatrixBase<Derived>::PlainMatrixType* result);
/** \ingroup MatrixFunctions_Module
@@ -90,13 +90,13 @@ template <typename MatrixType>
class MatrixExponential {
public:
/** \brief Compute the matrix exponential.
/** \brief Compute the matrix exponential.
*
* \param M matrix whose exponential is to be computed.
* \param M matrix whose exponential is to be computed.
* \param result pointer to the matrix in which to store the result.
*/
MatrixExponential(const MatrixType &M, MatrixType *result);
MatrixExponential(const MatrixType &M, MatrixType *result);
private:
@@ -105,7 +105,7 @@ class MatrixExponential {
MatrixExponential& operator=(const MatrixExponential&);
/** \brief Compute the (3,3)-Pad&eacute; approximant to the exponential.
*
*
* After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
* approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
*
@@ -114,7 +114,7 @@ class MatrixExponential {
void pade3(const MatrixType &A);
/** \brief Compute the (5,5)-Pad&eacute; approximant to the exponential.
*
*
* After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
* approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
*
@@ -123,7 +123,7 @@ class MatrixExponential {
void pade5(const MatrixType &A);
/** \brief Compute the (7,7)-Pad&eacute; approximant to the exponential.
*
*
* After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
* approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
*
@@ -132,7 +132,7 @@ class MatrixExponential {
void pade7(const MatrixType &A);
/** \brief Compute the (9,9)-Pad&eacute; approximant to the exponential.
*
*
* After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
* approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
*
@@ -141,7 +141,7 @@ class MatrixExponential {
void pade9(const MatrixType &A);
/** \brief Compute the (13,13)-Pad&eacute; approximant to the exponential.
*
*
* After exit, \f$ (V+U)(V-U)^{-1} \f$ is the Pad&eacute;
* approximant of \f$ \exp(A) \f$ around \f$ A = 0 \f$.
*
@@ -149,10 +149,10 @@ class MatrixExponential {
*/
void pade13(const MatrixType &A);
/** \brief Compute Pad&eacute; approximant to the exponential.
*
* Computes \c m_U, \c m_V and \c m_squarings such that
* \f$ (V+U)(V-U)^{-1} \f$ is a Pad&eacute; of
/** \brief Compute Pad&eacute; approximant to the exponential.
*
* Computes \c m_U, \c m_V and \c m_squarings such that
* \f$ (V+U)(V-U)^{-1} \f$ is a Pad&eacute; of
* \f$ \exp(2^{-\mbox{squarings}}M) \f$ around \f$ M = 0 \f$. The
* degree of the Pad&eacute; approximant and the value of
* squarings are chosen such that the approximation error is no
@@ -164,7 +164,7 @@ class MatrixExponential {
*/
void computeUV(double);
/** \brief Compute Pad&eacute; approximant to the exponential.
/** \brief Compute Pad&eacute; approximant to the exponential.
*
* \sa computeUV(double);
*/
@@ -174,7 +174,7 @@ class MatrixExponential {
typedef typename NumTraits<typename ei_traits<MatrixType>::Scalar>::Real RealScalar;
/** \brief Pointer to matrix whose exponential is to be computed. */
const MatrixType* m_M;
const MatrixType* m_M;
/** \brief Even-degree terms in numerator of Pad&eacute; approximant. */
MatrixType m_U;
@@ -200,14 +200,14 @@ class MatrixExponential {
template <typename MatrixType>
MatrixExponential<MatrixType>::MatrixExponential(const MatrixType &M, MatrixType *result) :
m_M(&M),
m_U(M.rows(),M.cols()),
m_V(M.rows(),M.cols()),
m_tmp1(M.rows(),M.cols()),
m_tmp2(M.rows(),M.cols()),
m_Id(MatrixType::Identity(M.rows(), M.cols())),
m_squarings(0),
m_l1norm(static_cast<float>(M.cwise().abs().colwise().sum().maxCoeff()))
m_M(&M),
m_U(M.rows(),M.cols()),
m_V(M.rows(),M.cols()),
m_tmp1(M.rows(),M.cols()),
m_tmp2(M.rows(),M.cols()),
m_Id(MatrixType::Identity(M.rows(), M.cols())),
m_squarings(0),
m_l1norm(static_cast<float>(M.cwiseAbs().colwise().sum().maxCoeff()))
{
computeUV(RealScalar());
m_tmp1 = m_U + m_V; // numerator of Pade approximant
@@ -267,8 +267,8 @@ EIGEN_STRONG_INLINE void MatrixExponential<MatrixType>::pade9(const MatrixType &
template <typename MatrixType>
EIGEN_STRONG_INLINE void MatrixExponential<MatrixType>::pade13(const MatrixType &A)
{
const Scalar b[] = {64764752532480000., 32382376266240000., 7771770303897600.,
1187353796428800., 129060195264000., 10559470521600., 670442572800.,
const Scalar b[] = {64764752532480000., 32382376266240000., 7771770303897600.,
1187353796428800., 129060195264000., 10559470521600., 670442572800.,
33522128640., 1323241920., 40840800., 960960., 16380., 182., 1.};
MatrixType A2 = A * A;
MatrixType A4 = A2 * A2;
@@ -317,7 +317,7 @@ void MatrixExponential<MatrixType>::computeUV(double)
}
template <typename Derived>
EIGEN_STRONG_INLINE void ei_matrix_exponential(const MatrixBase<Derived> &M,
EIGEN_STRONG_INLINE void ei_matrix_exponential(const MatrixBase<Derived> &M,
typename MatrixBase<Derived>::PlainMatrixType* result)
{
ei_assert(M.rows() == M.cols());

16
unsupported/Eigen/src/MatrixFunctions/MatrixFunctionAtomic.h
View File

@@ -25,7 +25,7 @@
#ifndef EIGEN_MATRIX_FUNCTION_ATOMIC
#define EIGEN_MATRIX_FUNCTION_ATOMIC
/** \ingroup MatrixFunctions_Module
/** \ingroup MatrixFunctions_Module
* \class MatrixFunctionAtomic
* \brief Helper class for computing matrix functions of atomic matrices.
*
@@ -110,30 +110,30 @@ void MatrixFunctionAtomic<MatrixType>::computeMu()
const MatrixType N = MatrixType::Identity(m_Arows, m_Arows) - m_Ashifted;
VectorType e = VectorType::Ones(m_Arows);
N.template triangularView<UpperTriangular>().solveInPlace(e);
m_mu = e.cwise().abs().maxCoeff();
m_mu = e.cwiseAbs().maxCoeff();
}
/** \brief Determine whether Taylor series has converged */
template <typename MatrixType>
bool MatrixFunctionAtomic<MatrixType>::taylorConverged(int s, const MatrixType& F,
bool MatrixFunctionAtomic<MatrixType>::taylorConverged(int s, const MatrixType& F,
const MatrixType& Fincr, const MatrixType& P)
{
const int n = F.rows();
const RealScalar F_norm = F.cwise().abs().rowwise().sum().maxCoeff();
const RealScalar Fincr_norm = Fincr.cwise().abs().rowwise().sum().maxCoeff();
const RealScalar F_norm = F.cwiseAbs().rowwise().sum().maxCoeff();
const RealScalar Fincr_norm = Fincr.cwiseAbs().rowwise().sum().maxCoeff();
if (Fincr_norm < epsilon<Scalar>() * F_norm) {
RealScalar delta = 0;
RealScalar rfactorial = 1;
for (int r = 0; r < n; r++) {
RealScalar mx = 0;
for (int i = 0; i < n; i++)
for (int i = 0; i < n; i++)
mx = std::max(mx, std::abs(m_f(m_Ashifted(i, i) + m_avgEival, s+r)));
if (r != 0)
rfactorial *= r;
delta = std::max(delta, mx / rfactorial);
}
const RealScalar P_norm = P.cwise().abs().rowwise().sum().maxCoeff();
if (m_mu * delta * P_norm < epsilon<Scalar>() * F_norm)
const RealScalar P_norm = P.cwiseAbs().rowwise().sum().maxCoeff();
if (m_mu * delta * P_norm < epsilon<Scalar>() * F_norm)
return true;
}
return false;

22
unsupported/Eigen/src/NonLinearOptimization/HybridNonLinearSolver.h
View File

@@ -36,7 +36,7 @@
*
* The user must provide a subroutine which calculates the
* functions. The Jacobian is either provided by the user, or approximated
* using a forward-difference method.
* using a forward-difference method.
*
*/
template<typename FunctorType, typename Scalar=double>
@@ -50,7 +50,7 @@ public:
Running = -1,
ImproperInputParameters = 0,
RelativeErrorTooSmall = 1,
TooManyFunctionEvaluation = 2,
TooManyFunctionEvaluation = 2,
TolTooSmall = 3,
NotMakingProgressJacobian = 4,
NotMakingProgressIterations = 5,
@@ -156,7 +156,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::hybrj1(
parameters.xtol = tol;
diag.setConstant(n, 1.);
return solve(
x,
x,
2
);
}
@@ -241,7 +241,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveOneStep(
/* on the first iteration, calculate the norm of the scaled x */
/* and initialize the step bound delta. */
wa3 = diag.cwise() * x;
wa3 = diag.cwiseProduct(x);
xnorm = wa3.stableNorm();
delta = parameters.factor * xnorm;
if (delta == 0.)
@@ -285,7 +285,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveOneStep(
/* Computing MAX */
if (mode != 2)
diag = diag.cwise().max(wa2);
diag = diag.cwiseMax(wa2);
/* beginning of the inner loop. */
@@ -299,7 +299,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveOneStep(
wa1 = -wa1;
wa2 = x + wa1;
wa3 = diag.cwise() * wa1;
wa3 = diag.cwiseProduct(wa1);
pnorm = wa3.stableNorm();
/* on the first iteration, adjust the initial step bound. */
@@ -364,7 +364,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveOneStep(
if (ratio >= Scalar(1e-4)) {
/* successful iteration. update x, fvec, and their norms. */
x = wa2;
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
fvec = wa4;
xnorm = wa2.stableNorm();
fnorm = fnorm1;
@@ -555,7 +555,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveNumericalDiffOneStep(
/* on the first iteration, calculate the norm of the scaled x */
/* and initialize the step bound delta. */
wa3 = diag.cwise() * x;
wa3 = diag.cwiseProduct(x);
xnorm = wa3.stableNorm();
delta = parameters.factor * xnorm;
if (delta == 0.)
@@ -599,7 +599,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveNumericalDiffOneStep(
/* Computing MAX */
if (mode != 2)
diag = diag.cwise().max(wa2);
diag = diag.cwiseMax(wa2);
/* beginning of the inner loop. */
@@ -613,7 +613,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveNumericalDiffOneStep(
wa1 = -wa1;
wa2 = x + wa1;
wa3 = diag.cwise() * wa1;
wa3 = diag.cwiseProduct(wa1);
pnorm = wa3.stableNorm();
/* on the first iteration, adjust the initial step bound. */
@@ -678,7 +678,7 @@ HybridNonLinearSolver<FunctorType,Scalar>::solveNumericalDiffOneStep(
if (ratio >= Scalar(1e-4)) {
/* successful iteration. update x, fvec, and their norms. */
x = wa2;
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
fvec = wa4;
xnorm = wa2.stableNorm();
fnorm = fnorm1;

26
unsupported/Eigen/src/NonLinearOptimization/LevenbergMarquardt.h
View File

@@ -37,7 +37,7 @@
* http://en.wikipedia.org/wiki/Levenberg%E2%80%93Marquardt_algorithm
*/
template<typename FunctorType, typename Scalar=double>
class LevenbergMarquardt
class LevenbergMarquardt
{
public:
LevenbergMarquardt(FunctorType &_functor)
@@ -50,7 +50,7 @@ public:
RelativeErrorTooSmall = 2,
RelativeErrorAndReductionTooSmall = 3,
CosinusTooSmall = 4,
TooManyFunctionEvaluation = 5,
TooManyFunctionEvaluation = 5,
FtolTooSmall = 6,
XtolTooSmall = 7,
GtolTooSmall = 8,
@@ -253,7 +253,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOneStep(
wa2 = fjac.colwise().blueNorm();
ei_qrfac<Scalar>(m, n, fjac.data(), fjac.rows(), true, ipvt.data(), wa1.data());
ipvt.cwise()-=1; // qrfac() creates ipvt with fortran convention (1->n), convert it to c (0->n-1)
ipvt.array() -= 1; // qrfac() creates ipvt with fortran convention (1->n), convert it to c (0->n-1)
/* on the first iteration and if mode is 1, scale according */
/* to the norms of the columns of the initial jacobian. */
@@ -269,7 +269,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOneStep(
/* on the first iteration, calculate the norm of the scaled x */
/* and initialize the step bound delta. */
wa3 = diag.cwise() * x;
wa3 = diag.cwiseProduct(x);
xnorm = wa3.stableNorm();
delta = parameters.factor * xnorm;
if (delta == 0.)
@@ -316,7 +316,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOneStep(
/* rescale if necessary. */
if (mode != 2) /* Computing MAX */
diag = diag.cwise().max(wa2);
diag = diag.cwiseMax(wa2);
/* beginning of the inner loop. */
do {
@@ -329,7 +329,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOneStep(
wa1 = -wa1;
wa2 = x + wa1;
wa3 = diag.cwise() * wa1;
wa3 = diag.cwiseProduct(wa1);
pnorm = wa3.stableNorm();
/* on the first iteration, adjust the initial step bound. */
@@ -395,7 +395,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOneStep(
if (ratio >= Scalar(1e-4)) {
/* successful iteration. update x, fvec, and their norms. */
x = wa2;
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
fvec = wa4;
xnorm = wa2.stableNorm();
fnorm = fnorm1;
@@ -538,10 +538,10 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOptimumStorageOneStep(
wa2[j] = fjac.col(j).head(j).stableNorm();
}
if (sing) {
ipvt.cwise()+=1;
ipvt.array() += 1;
wa2 = fjac.colwise().blueNorm();
ei_qrfac<Scalar>(n, n, fjac.data(), fjac.rows(), true, ipvt.data(), wa1.data());
ipvt.cwise()-=1; // qrfac() creates ipvt with fortran convention (1->n), convert it to c (0->n-1)
ipvt.array() -= 1; // qrfac() creates ipvt with fortran convention (1->n), convert it to c (0->n-1)
for (j = 0; j < n; ++j) {
if (fjac(j,j) != 0.) {
sum = 0.;
@@ -569,7 +569,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOptimumStorageOneStep(
/* on the first iteration, calculate the norm of the scaled x */
/* and initialize the step bound delta. */
wa3 = diag.cwise() * x;
wa3 = diag.cwiseProduct(x);
xnorm = wa3.stableNorm();
delta = parameters.factor * xnorm;
if (delta == 0.)
@@ -599,7 +599,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOptimumStorageOneStep(
/* rescale if necessary. */
if (mode != 2) /* Computing MAX */
diag = diag.cwise().max(wa2);
diag = diag.cwiseMax(wa2);
/* beginning of the inner loop. */
do {
@@ -612,7 +612,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOptimumStorageOneStep(
wa1 = -wa1;
wa2 = x + wa1;
wa3 = diag.cwise() * wa1;
wa3 = diag.cwiseProduct(wa1);
pnorm = wa3.stableNorm();
/* on the first iteration, adjust the initial step bound. */
@@ -678,7 +678,7 @@ LevenbergMarquardt<FunctorType,Scalar>::minimizeOptimumStorageOneStep(
if (ratio >= Scalar(1e-4)) {
/* successful iteration. update x, fvec, and their norms. */
x = wa2;
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
fvec = wa4;
xnorm = wa2.stableNorm();
fnorm = fnorm1;

4
unsupported/Eigen/src/NonLinearOptimization/dogleg.h
View File

@@ -50,7 +50,7 @@ void ei_dogleg(
/* test whether the gauss-newton direction is acceptable. */
wa1.fill(0.);
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
qnorm = wa2.stableNorm();
if (qnorm <= delta)
return;
@@ -80,7 +80,7 @@ void ei_dogleg(
/* calculate the point along the scaled gradient */
/* at which the quadratic is minimized. */
wa1.cwise() /= diag*gnorm;
wa1.array() /= (diag*gnorm).array();
l = 0;
for (j = 0; j < n; ++j) {
sum = 0.;

8
unsupported/Eigen/src/NonLinearOptimization/lmpar.h
View File

@@ -36,7 +36,7 @@ void ei_lmpar(
for (j = 0; j < n; ++j) {
if (r(j,j) == 0. && nsing == n-1)
nsing = j - 1;
if (nsing < n-1)
if (nsing < n-1)
wa1[j] = 0.;
}
for (j = nsing; j>=0; --j) {
@@ -54,7 +54,7 @@ void ei_lmpar(
/* for acceptance of the gauss-newton direction. */
iter = 0;
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
dxnorm = wa2.blueNorm();
fp = dxnorm - delta;
if (fp <= Scalar(0.1) * delta) {
@@ -76,7 +76,7 @@ void ei_lmpar(
// way:
for (j = 0; j < n; ++j) {
Scalar sum = 0.;
for (i = 0; i < j; ++i)
for (i = 0; i < j; ++i)
sum += r(i,j) * wa1[i];
wa1[j] = (wa1[j] - sum) / r(j,j);
}
@@ -117,7 +117,7 @@ void ei_lmpar(
Matrix< Scalar, Dynamic, 1 > sdiag(n);
ei_qrsolv<Scalar>(r, ipvt, wa1, qtb, x, sdiag);
wa2 = diag.cwise() * x;
wa2 = diag.cwiseProduct(x);
dxnorm = wa2.blueNorm();
temp = fp;
fp = dxnorm - delta;