fix sparse LDLT with complexes

This commit is contained in:
Gael Guennebaud 2010-06-03 11:56:08 +02:00
parent 8350ab9fb8
commit d92de9574a
3 changed files with 22 additions and 27 deletions

View File

@ -71,6 +71,8 @@ LDL License:
* *
* \param MatrixType the type of the matrix of which we are computing the LDLT Cholesky decomposition * \param MatrixType the type of the matrix of which we are computing the LDLT Cholesky decomposition
* *
* \warning the upper triangular part has to be specified. The rest of the matrix is not used. The input matrix must be column major.
*
* \sa class LDLT, class LDLT * \sa class LDLT, class LDLT
*/ */
template<typename MatrixType, int Backend = DefaultBackend> template<typename MatrixType, int Backend = DefaultBackend>
@ -280,7 +282,7 @@ bool SparseLDLT<MatrixType,Backend>::_numeric(const MatrixType& a)
Index i = Pinv ? Pinv[Ai[p]] : Ai[p]; /* get A(i,k) */ Index i = Pinv ? Pinv[Ai[p]] : Ai[p]; /* get A(i,k) */
if (i <= k) if (i <= k)
{ {
y[i] += Ax[p]; /* scatter A(i,k) into Y (sum duplicates) */ y[i] += ei_conj(Ax[p]); /* scatter A(i,k) into Y (sum duplicates) */
Index len; Index len;
for (len = 0; tags[i] != k; i = m_parent[i]) for (len = 0; tags[i] != k; i = m_parent[i])
{ {
@ -291,22 +293,23 @@ bool SparseLDLT<MatrixType,Backend>::_numeric(const MatrixType& a)
pattern[--top] = pattern[--len]; pattern[--top] = pattern[--len];
} }
} }
/* compute numerical values kth row of L (a sparse triangular solve) */ /* compute numerical values kth row of L (a sparse triangular solve) */
m_diag[k] = y[k]; /* get D(k,k) and clear Y(k) */ m_diag[k] = y[k]; /* get D(k,k) and clear Y(k) */
y[k] = 0.0; y[k] = 0.0;
for (; top < size; ++top) for (; top < size; ++top)
{ {
Index i = pattern[top]; /* pattern[top:n-1] is pattern of L(:,k) */ Index i = pattern[top]; /* pattern[top:n-1] is pattern of L(:,k) */
Scalar yi = y[i]; /* get and clear Y(i) */ Scalar yi = (y[i]); /* get and clear Y(i) */
y[i] = 0.0; y[i] = 0.0;
Index p2 = Lp[i] + m_nonZerosPerCol[i]; Index p2 = Lp[i] + m_nonZerosPerCol[i];
Index p; Index p;
for (p = Lp[i]; p < p2; ++p) for (p = Lp[i]; p < p2; ++p)
y[Li[p]] -= Lx[p] * yi; y[Li[p]] -= ei_conj(Lx[p]) * (yi);
Scalar l_ki = yi / m_diag[i]; /* the nonzero entry L(k,i) */ Scalar l_ki = yi / m_diag[i]; /* the nonzero entry L(k,i) */
m_diag[k] -= l_ki * yi; m_diag[k] -= l_ki * ei_conj(yi);
Li[p] = k; /* store L(k,i) in column form of L */ Li[p] = k; /* store L(k,i) in column form of L */
Lx[p] = l_ki; Lx[p] = (l_ki);
++m_nonZerosPerCol[i]; /* increment count of nonzeros in col i */ ++m_nonZerosPerCol[i]; /* increment count of nonzeros in col i */
} }
if (m_diag[k] == 0.0) if (m_diag[k] == 0.0)
@ -323,7 +326,7 @@ bool SparseLDLT<MatrixType,Backend>::_numeric(const MatrixType& a)
return ok; /* success, diagonal of D is all nonzero */ return ok; /* success, diagonal of D is all nonzero */
} }
/** Computes b = L^-T L^-1 b */ /** Computes b = L^-T D^-1 L^-1 b */
template<typename MatrixType, int Backend> template<typename MatrixType, int Backend>
template<typename Derived> template<typename Derived>
bool SparseLDLT<MatrixType, Backend>::solveInPlace(MatrixBase<Derived> &b) const bool SparseLDLT<MatrixType, Backend>::solveInPlace(MatrixBase<Derived> &b) const
@ -336,8 +339,6 @@ bool SparseLDLT<MatrixType, Backend>::solveInPlace(MatrixBase<Derived> &b) const
if (m_matrix.nonZeros()>0) // otherwise L==I if (m_matrix.nonZeros()>0) // otherwise L==I
m_matrix.template triangularView<UnitLower>().solveInPlace(b); m_matrix.template triangularView<UnitLower>().solveInPlace(b);
b = b.cwiseQuotient(m_diag); b = b.cwiseQuotient(m_diag);
// FIXME should be .adjoint() but it fails to compile...
if (m_matrix.nonZeros()>0) // otherwise L==I if (m_matrix.nonZeros()>0) // otherwise L==I
m_matrix.adjoint().template triangularView<UnitUpper>().solveInPlace(b); m_matrix.adjoint().template triangularView<UnitUpper>().solveInPlace(b);

View File

@ -195,14 +195,7 @@ bool SparseLLT<MatrixType, Backend>::solveInPlace(MatrixBase<Derived> &b) const
ei_assert(size==b.rows()); ei_assert(size==b.rows());
m_matrix.template triangularView<Lower>().solveInPlace(b); m_matrix.template triangularView<Lower>().solveInPlace(b);
// FIXME should be simply .adjoint() but it fails to compile... m_matrix.adjoint().template triangularView<Upper>().solveInPlace(b);
if (NumTraits<Scalar>::IsComplex)
{
CholMatrixType aux = m_matrix.conjugate();
aux.transpose().template triangularView<Upper>().solveInPlace(b);
}
else
m_matrix.transpose().template triangularView<Upper>().solveInPlace(b);
return true; return true;
} }

View File

@ -149,26 +149,27 @@ template<typename Scalar> void sparse_solvers(int rows, int cols)
} }
// test LDLT // test LDLT
if (!NumTraits<Scalar>::IsComplex)
{ {
// TODO fix the issue with complex (see SparseLDLT::solveInPlace)
SparseMatrix<Scalar> m2(rows, cols); SparseMatrix<Scalar> m2(rows, cols);
DenseMatrix refMat2(rows, cols); DenseMatrix refMat2(rows, cols);
DenseVector b = DenseVector::Random(cols); DenseVector b = DenseVector::Random(cols);
DenseVector refX(cols), x(cols); DenseVector refX(cols), x(cols);
// initSPD(density, refMat2, m2);
initSparse<Scalar>(density, refMat2, m2, ForceNonZeroDiag|MakeUpperTriangular, 0, 0); initSparse<Scalar>(density, refMat2, m2, ForceNonZeroDiag|MakeUpperTriangular, 0, 0);
refMat2 += (refMat2.adjoint()).eval(); for(int i=0; i<rows; ++i)
refMat2.diagonal() *= 0.5; m2.coeffRef(i,i) = refMat2(i,i) = ei_abs(ei_real(refMat2(i,i)));
refX = refMat2.llt().solve(b); // FIXME use LLT to compute the reference because LDLT seems to fail with large matrices refX = refMat2.template selfadjointView<Upper>().llt().solve(b);
// FIXME use LLT to compute the reference because LDLT seems to fail with large matrices
typedef SparseMatrix<Scalar,Upper|SelfAdjoint> SparseSelfAdjointMatrix; typedef SparseMatrix<Scalar,Upper|SelfAdjoint> SparseSelfAdjointMatrix;
x = b; x = b;
SparseLDLT<SparseSelfAdjointMatrix> ldlt(m2); SparseLDLT<SparseSelfAdjointMatrix> ldlt(m2);
if (ldlt.succeeded()) if (ldlt.succeeded())
ldlt.solveInPlace(x); ldlt.solveInPlace(x);
else
std::cerr << "warning LDLT failed\n";
VERIFY(refX.isApprox(x,test_precision<Scalar>()) && "LDLT: default"); VERIFY(refX.isApprox(x,test_precision<Scalar>()) && "LDLT: default");
} }