/*========================================================================= * * Copyright NumFOCUS * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * https://www.apache.org/licenses/LICENSE-2.0.txt * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. * *=========================================================================*/ #ifndef VNLIterativeSparseSolverTraits_h #define VNLIterativeSparseSolverTraits_h #include "vnl/vnl_vector.h" #include "vnl/vnl_sparse_matrix.h" #include "vnl/vnl_sparse_matrix_linear_system.h" #include "vnl/algo/vnl_lsqr.h" /** \class VNLIterativeSparseSolverTraits * \brief Generic interface for iterative sparse linear solver. * * This generic interface (common to several sparse solvers), allow to * interchange solver solutions when dealing with sparse linear system. See * itk::ParameterizationQuadEdgeMeshFilter for reference. * * It internally uses the VNL library to represent and deal with vectors * (vnl_vector) and sparse matrices (vnl_sparse_matrix). The solver by itself * is an iterative one see vnl_sparse_matrix_linear_system for more details on * the method used. * * \ingroup ITKCommon * * \sa VNLSparseLUSolverTraits */ template class VNLIterativeSparseSolverTraits { public: using ValueType = T; using MatrixType = vnl_sparse_matrix; using VectorType = vnl_vector; using SolverType = vnl_lsqr; /** \return false (it is not a direct solver, it is an iterative solver) */ static bool IsDirectSolver() { return false; } /** \brief initialize a square sparse matrix of size iN x iN */ static MatrixType InitializeSparseMatrix(const unsigned int & iN) { return MatrixType(iN, iN); } /** \brief initialize a sparse matrix of size iRow x iCol */ static MatrixType InitializeSparseMatrix(const unsigned int & iRow, const unsigned int & iCol) { return MatrixType(iRow, iCol); } /** \brief initialize a vector of size iN */ static VectorType InitializeVector(const unsigned int & iN) { return VectorType(iN); } /** \brief iA[iR][iC] = iV */ static void FillMatrix(MatrixType & iA, const unsigned int & iR, const unsigned int & iC, const ValueType & iV) { iA(iR, iC) = iV; } /** \brief iA[iR][iC] += iV */ static void AddToMatrix(MatrixType & iA, const unsigned int & iR, const unsigned int & iC, const ValueType & iV) { iA(iR, iC) += iV; } /** \brief Solve the linear system \f$ iA \cdot oX = iB \f$ */ static bool Solve(const MatrixType & iA, const VectorType & iB, VectorType & oX) { using SparseLinearSystemType = vnl_sparse_matrix_linear_system; SparseLinearSystemType system(iA, iB); SolverType solver(system); return solver.minimize(oX); } /** \brief Solve the linear systems: \f$ iA \cdot oX = iBx \f$, \f$ iA \cdot oY = iBy \f$, \f$ iA \cdot oZ = iBz \f$ */ static bool Solve(const MatrixType & iA, const VectorType & iBx, const VectorType & iBy, const VectorType & iBz, VectorType & oX, VectorType & oY, VectorType & oZ) { bool result1 = Solve(iA, iBx, 100000, oX); bool result2 = Solve(iA, iBy, 100000, oY); bool result3 = Solve(iA, iBz, 100000, oZ); return (result1 && result2 && result3); } /** \brief Solve the linear systems: \f$ iA \cdot oX = iBx \f$, \f$ iA \cdot oY = iBy \f$ */ static bool Solve(const MatrixType & iA, const VectorType & iBx, const VectorType & iBy, VectorType & oX, VectorType & oY) { bool result1 = Solve(iA, iBx, oX); bool result2 = Solve(iA, iBy, oY); return (result1 && result2); } /** \brief Solve the linear systems: \f$ iA \cdot oX = iBx \f$ in N iterations */ static bool Solve(const MatrixType & iA, const VectorType & iB, const long & iNbIter, VectorType & oX) { using SparseLinearSystemType = vnl_sparse_matrix_linear_system; SparseLinearSystemType system(iA, iB); SolverType solver(system); solver.set_max_iterations(iNbIter); return solver.minimize(oX); } }; #endif