vcglib/vcg/math/lin_algebra.h

630 lines
17 KiB
C++
Raw Blame History

/****************************************************************************
* VCGLib o o *
* Visual and Computer Graphics Library o o *
* _ O _ *
* Copyright(C) 2006 \/)\/ *
* Visual Computing Lab /\/| *
* ISTI - Italian National Research Council | *
* \ *
* All rights reserved. *
* *
* This program is free software; 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. *
* *
* This program 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 General Public License (http://www.gnu.org/licenses/gpl.txt) *
* for more details. *
* *
****************************************************************************/
/****************************************************************************
History
$Log: not supported by cvs2svn $
Revision 1.12 2006/07/24 07:26:47 fiorin
Changed the template argument in JacobiRotate and added method for sorting eigenvalues and eigenvectors (SortEigenvaluesAndEigenvectors)
Revision 1.11 2006/05/25 09:35:55 cignoni
added missing internal prototype to Sort function
Revision 1.10 2006/05/17 09:26:35 cignoni
Added initial disclaimer
****************************************************************************/
#ifndef __VCGLIB_LINALGEBRA_H
#define __VCGLIB_LINALGEBRA_H
#include <vcg/math/matrix44.h>
namespace vcg
{
/** \addtogroup math */
/* @{ */
/*!
*
*/
template< typename MATRIX_TYPE >
static void JacobiRotate(MATRIX_TYPE &A, typename MATRIX_TYPE::ScalarType s, typename MATRIX_TYPE::ScalarType tau, int i,int j,int k,int l)
{
typename MATRIX_TYPE::ScalarType g=A[i][j];
typename MATRIX_TYPE::ScalarType h=A[k][l];
A[i][j]=g-s*(h+g*tau);
A[k][l]=h+s*(g-h*tau);
};
/*!
* Computes all eigenvalues and eigenvectors of a real symmetric matrix .
* On output, elements of the input matrix above the diagonal are destroyed.
* \param d returns the eigenvalues of a.
* \param v is a matrix whose columns contain, the normalized eigenvectors
* \param nrot returns the number of Jacobi rotations that were required.
*/
template <typename MATRIX_TYPE, typename POINT_TYPE>
static void Jacobi(MATRIX_TYPE &w, POINT_TYPE &d, MATRIX_TYPE &v, int &nrot)
{
assert(w.RowsNumber()==w.ColumnsNumber());
int dimension = w.RowsNumber();
int j,iq,ip,i;
//assert(w.IsSymmetric());
typename MATRIX_TYPE::ScalarType tresh, theta, tau, t, sm, s, h, g, c;
POINT_TYPE b, z;
v.SetIdentity();
for (ip=0;ip<dimension;++ip) //Initialize b and d to the diagonal of a.
{
b[ip]=d[ip]=w[ip][ip];
z[ip]=0.0; //This vector will accumulate terms of the form tapq as in equation (11.1.14).
}
nrot=0;
for (i=0;i<50;i++)
{
sm=0.0;
for (ip=0;ip<dimension-1;++ip) // Sum off diagonal elements
{
for (iq=ip+1;iq<dimension;++iq)
sm += fabs(w[ip][iq]);
}
if (sm == 0.0) //The normal return, which relies on quadratic convergence to machine underflow.
{
return;
}
if (i < 4)
tresh=0.2*sm/(dimension*dimension); //...on the first three sweeps.
else
tresh=0.0; //...thereafter.
for (ip=0;ip<dimension-1;++ip)
{
for (iq=ip+1;iq<dimension;iq++)
{
g=100.0*fabs(w[ip][iq]);
//After four sweeps, skip the rotation if the off-diagonal element is small.
if(i>4 && (float)(fabs(d[ip])+g) == (float)fabs(d[ip]) && (float)(fabs(d[iq])+g) == (float)fabs(d[iq]))
w[ip][iq]=0.0;
else if (fabs(w[ip][iq]) > tresh)
{
h=d[iq]-d[ip];
if ((float)(fabs(h)+g) == (float)fabs(h))
t=(w[ip][iq])/h; //t =1/(2#)
else
{
theta=0.5*h/(w[ip][iq]); //Equation (11.1.10).
t=1.0/(fabs(theta)+sqrt(1.0+theta*theta));
if (theta < 0.0) t = -t;
}
c=1.0/sqrt(1+t*t);
s=t*c;
tau=s/(1.0+c);
h=t*w[ip][iq];
z[ip] -= h;
z[iq] += h;
d[ip] -= h;
d[iq] += h;
w[ip][iq]=0.0;
for (j=0;j<=ip-1;j++) { //Case of rotations 1 <= j < p.
JacobiRotate<MATRIX_TYPE>(w,s,tau,j,ip,j,iq) ;
}
for (j=ip+1;j<=iq-1;j++) { //Case of rotations p < j < q.
JacobiRotate<MATRIX_TYPE>(w,s,tau,ip,j,j,iq);
}
for (j=iq+1;j<dimension;j++) { //Case of rotations q< j <= n.
JacobiRotate<MATRIX_TYPE>(w,s,tau,ip,j,iq,j);
}
for (j=0;j<dimension;j++) {
JacobiRotate<MATRIX_TYPE>(v,s,tau,j,ip,j,iq);
}
++nrot;
}
}
}
for (ip=0;ip<dimension;ip++)
{
b[ip] += z[ip];
d[ip]=b[ip]; //Update d with the sum of ta_pq ,
z[ip]=0.0; //and reinitialize z.
}
}
};
/*!
* Given the eigenvectors and the eigenvalues as output from JacobiRotate, sorts the eigenvalues
* into descending order, and rearranges the columns of v correspondinlgy.
/param eigenvalues
/param eigenvector
*/
template < typename MATRIX_TYPE, typename POINT_TYPE >
void SortEigenvaluesAndEigenvectors(POINT_TYPE &eigenvalues, MATRIX_TYPE &eigenvectors)
{
assert(eigenvectors.ColumnsNumber()==eigenvectors.RowsNumber());
int dimension = eigenvectors.ColumnsNumber();
int i, j, k;
float p;
for (i=0; i<dimension-1; i++)
{
p = eigenvalues[ k=i ];
for (j=i+1; j<dimension; j++)
if (eigenvalues[j] >= p)
p = eigenvalues[ k=j ];
if (k != i)
{
eigenvalues[k] = eigenvalues[i]; // i.e.
eigenvalues[i] = p; // swaps the value of the elements i-th and k-th
for (j=0; j<dimension; j++)
{
p = eigenvectors[j][i]; // i.e.
eigenvectors[j][i] = eigenvectors[j][k]; // swaps the eigenvectors stored in the
eigenvectors[j][k] = p; // i-th and the k-th column
}
}
}
};
// Computes (a^2 + b^2)^(1/2) without destructive underflow or overflow.
template <typename TYPE>
inline static TYPE pythagora(TYPE a, TYPE b)
{
TYPE abs_a = fabs(a);
TYPE abs_b = fabs(b);
if (abs_a > abs_b)
return abs_a*sqrt(1.0+sqr(abs_b/abs_a));
else
return (abs_b == 0.0 ? 0.0 : abs_b*sqrt(1.0+sqr(abs_a/abs_b)));
};
template <typename TYPE>
inline static TYPE sign(TYPE a, TYPE b)
{
return (b >= 0.0 ? fabs(a) : -fabs(a));
};
template <typename TYPE>
inline static TYPE sqr(TYPE a)
{
TYPE sqr_arg = a;
return (sqr_arg == 0 ? 0 : sqr_arg*sqr_arg);
}
/*!
*
*/
enum SortingStrategy {LeaveUnsorted=0, SortAscending=1, SortDescending=2};
template< typename MATRIX_TYPE >
void Sort(MATRIX_TYPE &U, typename MATRIX_TYPE::ScalarType W[], MATRIX_TYPE &V, const SortingStrategy sorting) ;
/*!
* Given a matrix <I>A<SUB>m<>n</SUB></I>, this routine computes its singular value decomposition,
* i.e. <I>A=U<>W<EFBFBD>V<SUP>T</SUP></I>. The matrix <I>A</I> will be destroyed!
* (This is the implementation described in <I>Numerical Recipies</I>).
* \param A the matrix to be decomposed
* \param W the diagonal matrix of singular values <I>W</I>, stored as a vector <I>W[1...N]</I>
* \param V the matrix <I>V</I> (not the transpose <I>V<SUP>T</SUP></I>)
* \param max_iters max iteration number (default = 30).
* \return
*/
template <typename MATRIX_TYPE>
static bool SingularValueDecomposition(MATRIX_TYPE &A, typename MATRIX_TYPE::ScalarType *W, MATRIX_TYPE &V, const SortingStrategy sorting=LeaveUnsorted, const int max_iters=30)
{
typedef typename MATRIX_TYPE::ScalarType ScalarType;
int m = (int) A.RowsNumber();
int n = (int) A.ColumnsNumber();
int flag,i,its,j,jj,k,l,nm;
double anorm, c, f, g, h, s, scale, x, y, z, *rv1;
bool convergence = true;
rv1 = new double[n];
g = scale = anorm = 0;
// Householder reduction to bidiagonal form.
for (i=0; i<n; i++)
{
l = i+1;
rv1[i] = scale*g;
g = s = scale = 0.0;
if (i < m)
{
for (k = i; k<m; k++)
scale += fabs(A[k][i]);
if (scale)
{
for (k=i; k<m; k++)
{
A[k][i] /= scale;
s += A[k][i]*A[k][i];
}
f=A[i][i];
g = -sign<double>( sqrt(s), f );
h = f*g - s;
A[i][i]=f-g;
for (j=l; j<n; j++)
{
for (s=0.0, k=i; k<m; k++)
s += A[k][i]*A[k][j];
f = s/h;
for (k=i; k<m; k++)
A[k][j] += f*A[k][i];
}
for (k=i; k<m; k++)
A[k][i] *= scale;
}
}
W[i] = scale *g;
g = s = scale = 0.0;
if (i < m && i != (n-1))
{
for (k=l; k<n; k++)
scale += fabs(A[i][k]);
if (scale)
{
for (k=l; k<n; k++)
{
A[i][k] /= scale;
s += A[i][k]*A[i][k];
}
f = A[i][l];
g = -sign<double>(sqrt(s),f);
h = f*g - s;
A[i][l] = f-g;
for (k=l; k<n; k++)
rv1[k] = A[i][k]/h;
for (j=l; j<m; j++)
{
for (s=0.0, k=l; k<n; k++)
s += A[j][k]*A[i][k];
for (k=l; k<n; k++)
A[j][k] += s*rv1[k];
}
for (k=l; k<n; k++)
A[i][k] *= scale;
}
}
anorm=math::Max( anorm, (fabs(W[i])+fabs(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)
{
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++)
{
for (s=0.0, k=l; k<n; k++)
s += A[i][k] * V[k][j];
for (k=l; k<n; k++)
V[k][j] += s*V[k][i];
}
}
for (j=l; j<n; j++)
V[i][j] = V[j][i] = 0.0;
}
V[i][i] = 1.0;
g = rv1[i];
l = i;
}
// Accumulation of left-hand transformations.
for (i=math::Min(m,n)-1; i>=0; i--)
{
l = i+1;
g = W[i];
for (j=l; j<n; j++)
A[i][j]=0.0;
if (g)
{
g = 1.0/g;
for (j=l; j<n; j++)
{
for (s=0.0, k=l; k<m; k++)
s += A[k][i]*A[k][j];
f = (s/A[i][i])*g;
for (k=i; k<m; k++)
A[k][j] += f*A[k][i];
}
for (j=i; j<m; j++)
A[j][i] *= g;
}
else
for (j=i; j<m; j++)
A[j][i] = 0.0;
++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=1; its<=max_iters; its++)
{
flag=1;
for (l=k; l>=0; l--)
{
// Test for splitting.
nm=l-1;
// Note that rv1[1] is always zero.
if ((double)(fabs(rv1[l])+anorm) == anorm)
{
flag=0;
break;
}
if ((double)(fabs(W[nm])+anorm) == anorm)
break;
}
if (flag)
{
c=0.0; //Cancellation of rv1[l], if l > 1.
s=1.0;
for (i=l ;i<=k; i++)
{
f = s*rv1[i];
rv1[i] = c*rv1[i];
if ((double)(fabs(f)+anorm) == anorm)
break;
g = W[i];
h = pythagora<double>(f,g);
W[i] = h;
h = 1.0/h;
c = g*h;
s = -f*h;
for (j=0; j<m; j++)
{
y = A[j][nm];
z = A[j][i];
A[j][nm] = y*c + z*s;
A[j][i] = z*c - y*s;
}
}
}
z = W[k];
if (l == k) //Convergence.
{
if (z < 0.0) { // Singular value is made nonnegative.
W[k] = -z;
for (j=0; j<n; j++)
V[j][k] = -V[j][k];
}
break;
}
if (its == max_iters)
{
printf("no convergence in %d SingularValueDecomposition iterations\n", 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))/(2.0*h*y);
g = pythagora<double>(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 = pythagora<double>(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;
for (jj=0; jj<n; jj++)
{
x = V[jj][j];
z = V[jj][i];
V[jj][j] = x*c + z*s;
V[jj][i] = z*c - x*s;
}
z = pythagora<double>(f,h);
W[j] = z;
// Rotation can be arbitrary if z = 0.
if (z)
{
z = 1.0/z;
c = f*z;
s = h*z;
}
f = c*g + s*y;
x = c*y - s*g;
for (jj=0; jj<m; jj++)
{
y = A[jj][j];
z = A[jj][i];
A[jj][j] = y*c + z*s;
A[jj][i] = z*c - y*s;
}
}
rv1[l] = 0.0;
rv1[k] = f;
W[k] = x;
}
}
delete []rv1;
if (sorting!=LeaveUnsorted)
Sort<MATRIX_TYPE>(A, W, V, sorting);
return convergence;
};
/*!
* Sort the singular values computed by the <CODE>SingularValueDecomposition</CODE> procedure and
* modify the matrices <I>U</I> and <I>V</I> accordingly.
*/
// TODO modify the last parameter type
template< typename MATRIX_TYPE >
void Sort(MATRIX_TYPE &U, typename MATRIX_TYPE::ScalarType W[], MATRIX_TYPE &V, const SortingStrategy sorting)
{
typedef typename MATRIX_TYPE::ScalarType ScalarType;
assert(U.ColumnsNumber()==V.ColumnsNumber());
int mu = U.RowsNumber();
int mv = V.RowsNumber();
int n = U.ColumnsNumber();
//ScalarType* u = &U[0][0];
//ScalarType* v = &V[0][0];
for (int i=0; i<n; i++)
{
int k = i;
ScalarType p = W[i];
switch (sorting)
{
case SortAscending:
{
for (int j=i+1; j<n; j++)
{
if (W[j] < p)
{
k = j;
p = W[j];
}
}
break;
}
case SortDescending:
{
for (int j=i+1; j<n; j++)
{
if (W[j] > p)
{
k = j;
p = W[j];
}
}
break;
}
}
if (k != i)
{
W[k] = W[i]; // i.e.
W[i] = p; // swaps the i-th and the k-th elements
int j = mu;
//ScalarType* uji = u + i; // uji = &U[0][i]
//ScalarType* ujk = u + k; // ujk = &U[0][k]
//ScalarType* vji = v + i; // vji = &V[0][i]
//ScalarType* vjk = v + k; // vjk = &V[0][k]
//if (j)
//{
// for(;;) for( ; j!=0; --j, uji+=n, ujk+=n)
// { {
// p = *uji; p = *uji; // i.e.
// *uji = *ujk; *uji = *ujk; // swap( U[s][i], U[s][k] )
// *ujk = p; *ujk = p; //
// if (!(--j)) }
// break;
// uji += n;
// ujk += n;
// }
//}
for(int s=0; j!=0; ++s, --j)
std::swap(U[s][i], U[s][k]);
j = mv;
//if (j!=0)
//{
// for(;;) for ( ; j!=0; --j, vji+=n, ujk+=n)
// { {
// p = *vji; p = *vji; // i.e.
// *vji = *vjk; *vji = *vjk; // swap( V[s][i], V[s][k] )
// *vjk = p; *vjk = p; //
// if (!(--j)) }
// break;
// vji += n;
// vjk += n;
// }
//}
for (int s=0; j!=0; ++s, --j)
std::swap(V[s][i], V[s][k]);
}
}
}
/*!
* Solves A<>X = B for a vector X, where A is specified by the matrices <I>U<SUB>m<>n</SUB></I>,
* <I>W<SUB>n<>1</SUB></I> and <I>V<SUB>n<>n</SUB></I> as returned by <CODE>SingularValueDecomposition</CODE>.
* No input quantities are destroyed, so the routine may be called sequentially with different b<>s.
* \param x is the output solution vector (<I>x<SUB>n<>1</SUB></I>)
* \param b is the input right-hand side (<I>b<SUB>n<>1</SUB></I>)
*/
template <typename MATRIX_TYPE>
static void SingularValueBacksubstitution(const MATRIX_TYPE &U,
const typename MATRIX_TYPE::ScalarType *W,
const MATRIX_TYPE &V,
typename MATRIX_TYPE::ScalarType *x,
const typename MATRIX_TYPE::ScalarType *b)
{
typedef typename MATRIX_TYPE::ScalarType ScalarType;
unsigned int jj, j, i;
unsigned int columns_number = U.ColumnsNumber();
unsigned int rows_number = U.RowsNumber();
ScalarType s;
ScalarType *tmp = new ScalarType[columns_number];
for (j=0; j<columns_number; j++) //Calculate U^T * B.
{
s = 0;
if (W[j]!=0) //Nonzero result only if wj is nonzero.
{
for (i=0; i<rows_number; i++)
s += U[i][j]*b[i];
s /= W[j]; //This is the divide by wj .
}
tmp[j]=s;
}
for (j=0;j<columns_number;j++) //Matrix multiply by V to get answer.
{
s = 0;
for (jj=0; jj<columns_number; jj++)
s += V[j][jj]*tmp[jj];
x[j]=s;
}
delete []tmp;
};
/*! @} */
}; // end of namespace
#endif