1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
|
/*
* Copyright © 2004-2011 Ondra Kamenik
* Copyright © 2019 Dynare Team
*
* This file is part of Dynare.
*
* Dynare 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 3 of the License, or
* (at your option) any later version.
*
* Dynare 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 for more details.
*
* You should have received a copy of the GNU General Public License
* along with Dynare. If not, see <http://www.gnu.org/licenses/>.
*/
#include "SimilarityDecomp.hh"
#include "SchurDecomp.hh"
#include "SchurDecompEig.hh"
#include "SylvException.hh"
#include <dynlapack.h>
#include <cmath>
SimilarityDecomp::SimilarityDecomp(const ConstVector &d, int d_size, double log10norm)
{
SchurDecomp sd(SqSylvMatrix(Vector{d}, d_size));
q = std::make_unique<SqSylvMatrix>(sd.getQ());
b = std::make_unique<BlockDiagonal>(sd.getT());
invq = std::make_unique<SqSylvMatrix>(d_size);
invq->setUnit();
invq->multLeftTrans(sd.getQ());
double norm = pow(10.0, log10norm);
diagonalize(norm);
}
void
SimilarityDecomp::getXDim(diag_iter start, diag_iter end,
int &rows, int &cols) const
{
int si = start->getIndex();
int ei = end->getIndex();
cols = b->nrows() - ei;
rows = ei - si;
}
/* Find solution of X for diagonal block given by start(incl.) and
end(excl.). If the solution cannot be found, or it is greater than
norm, X is not changed and flase is returned.
*/
bool
SimilarityDecomp::solveX(diag_iter start, diag_iter end,
GeneralMatrix &X, double norm) const
{
int si = start->getIndex();
int ei = end->getIndex();
SqSylvMatrix A(const_cast<const BlockDiagonal &>(*b), si, si, X.nrows());
SqSylvMatrix B(const_cast<const BlockDiagonal &>(*b), ei, ei, X.ncols());
GeneralMatrix C(const_cast<const BlockDiagonal &>(*b), si, ei, X.nrows(), X.ncols());
lapack_int isgn = -1;
lapack_int m = A.nrows();
lapack_int n = B.nrows();
lapack_int lda = A.getLD(), ldb = B.getLD();
double scale;
lapack_int info;
dtrsyl("N", "N", &isgn, &m, &n, A.base(), &lda, B.base(), &ldb,
C.base(), &m, &scale, &info);
if (info < -1)
throw SYLV_MES_EXCEPTION("Wrong parameter to LAPACK dtrsyl.");
if (info == 1 || scale < 1)
return false;
if (C.getData().getMax() > norm)
return false;
X = C;
return true;
}
/* ⎛I −X⎞ ⎛I X⎞
Multiply Q and invQ with ⎝0 I⎠ and ⎝0 I⎠ respectively. This also sets X=−X. */
void
SimilarityDecomp::updateTransform(diag_iter start, diag_iter end,
GeneralMatrix &X)
{
int si = start->getIndex();
int ei = end->getIndex();
SqSylvMatrix iX(q->nrows());
iX.setUnit();
iX.place(X, si, ei);
invq->GeneralMatrix::multLeft(iX);
iX.setUnit();
X.mult(-1.0);
iX.place(X, si, ei);
q->multRight(iX);
}
void
SimilarityDecomp::bringGuiltyBlock(diag_iter start, diag_iter &end)
{
double av = b->getAverageDiagSize(start, end);
diag_iter guilty = b->findClosestDiagBlock(end, b->diag_end(), av);
SchurDecompEig sd(*b); // works on b including diagonal structure
end = sd.bubbleEigen(guilty, end); // iterators are valid
++end;
q->multRight(sd.getQ());
invq->multLeftTrans(sd.getQ());
}
void
SimilarityDecomp::diagonalize(double norm)
{
diag_iter start = b->diag_begin();
diag_iter end = start;
++end;
while (end != b->diag_end())
{
int xrows;
int xcols;
getXDim(start, end, xrows, xcols);
GeneralMatrix X(xrows, xcols);
if (solveX(start, end, X, norm))
{
updateTransform(start, end, X);
b->setZeroBlockEdge(end);
start = end;
++end;
}
else
bringGuiltyBlock(start, end); // moves with end
}
}
void
SimilarityDecomp::check(SylvParams &pars, const GeneralMatrix &m) const
{
// M − Q·B·Q⁻¹
SqSylvMatrix c(getQ() * getB());
c.multRight(getInvQ());
c.add(-1.0, m);
pars.f_err1 = c.getNorm1();
pars.f_errI = c.getNormInf();
// I − Q·Q⁻¹
c.setUnit();
c.mult(-1);
c.multAndAdd(getQ(), getInvQ());
pars.viv_err1 = c.getNorm1();
pars.viv_errI = c.getNormInf();
// I − Q⁻¹·Q
c.setUnit();
c.mult(-1);
c.multAndAdd(getInvQ(), getQ());
pars.ivv_err1 = c.getNorm1();
pars.ivv_errI = c.getNormInf();
}
void
SimilarityDecomp::infoToPars(SylvParams &pars) const
{
pars.f_blocks = getB().getNumBlocks();
pars.f_largest = getB().getLargestBlock();
pars.f_zeros = getB().getNumZeros();
pars.f_offdiag = getB().getNumOffdiagonal();
}
|