Brian Silverman | 72890c2 | 2015-09-19 14:37:37 -0400 | [diff] [blame] | 1 | // This file is part of Eigen, a lightweight C++ template library |
| 2 | // for linear algebra. |
| 3 | // |
| 4 | // Copyright (C) 2008 Gael Guennebaud <gael.guennebaud@inria.fr> |
| 5 | |
| 6 | /* NOTE The functions of this file have been adapted from the GMM++ library */ |
| 7 | |
| 8 | //======================================================================== |
| 9 | // |
| 10 | // Copyright (C) 2002-2007 Yves Renard |
| 11 | // |
| 12 | // This file is a part of GETFEM++ |
| 13 | // |
| 14 | // Getfem++ is free software; you can redistribute it and/or modify |
| 15 | // it under the terms of the GNU Lesser General Public License as |
| 16 | // published by the Free Software Foundation; version 2.1 of the License. |
| 17 | // |
| 18 | // This program is distributed in the hope that it will be useful, |
| 19 | // but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 20 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 21 | // GNU Lesser General Public License for more details. |
| 22 | // You should have received a copy of the GNU Lesser General Public |
| 23 | // License along with this program; if not, write to the Free Software |
| 24 | // Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, |
| 25 | // USA. |
| 26 | // |
| 27 | //======================================================================== |
| 28 | |
| 29 | #include "../../../../Eigen/src/Core/util/NonMPL2.h" |
| 30 | |
| 31 | #ifndef EIGEN_CONSTRAINEDCG_H |
| 32 | #define EIGEN_CONSTRAINEDCG_H |
| 33 | |
| 34 | #include <Eigen/Core> |
| 35 | |
| 36 | namespace Eigen { |
| 37 | |
| 38 | namespace internal { |
| 39 | |
| 40 | /** \ingroup IterativeSolvers_Module |
| 41 | * Compute the pseudo inverse of the non-square matrix C such that |
| 42 | * \f$ CINV = (C * C^T)^{-1} * C \f$ based on a conjugate gradient method. |
| 43 | * |
| 44 | * This function is internally used by constrained_cg. |
| 45 | */ |
| 46 | template <typename CMatrix, typename CINVMatrix> |
| 47 | void pseudo_inverse(const CMatrix &C, CINVMatrix &CINV) |
| 48 | { |
| 49 | // optimisable : copie de la ligne, precalcul de C * trans(C). |
| 50 | typedef typename CMatrix::Scalar Scalar; |
| 51 | typedef typename CMatrix::Index Index; |
| 52 | // FIXME use sparse vectors ? |
| 53 | typedef Matrix<Scalar,Dynamic,1> TmpVec; |
| 54 | |
| 55 | Index rows = C.rows(), cols = C.cols(); |
| 56 | |
| 57 | TmpVec d(rows), e(rows), l(cols), p(rows), q(rows), r(rows); |
| 58 | Scalar rho, rho_1, alpha; |
| 59 | d.setZero(); |
| 60 | |
| 61 | typedef Triplet<double> T; |
| 62 | std::vector<T> tripletList; |
| 63 | |
| 64 | for (Index i = 0; i < rows; ++i) |
| 65 | { |
| 66 | d[i] = 1.0; |
| 67 | rho = 1.0; |
| 68 | e.setZero(); |
| 69 | r = d; |
| 70 | p = d; |
| 71 | |
| 72 | while (rho >= 1e-38) |
| 73 | { /* conjugate gradient to compute e */ |
| 74 | /* which is the i-th row of inv(C * trans(C)) */ |
| 75 | l = C.transpose() * p; |
| 76 | q = C * l; |
| 77 | alpha = rho / p.dot(q); |
| 78 | e += alpha * p; |
| 79 | r += -alpha * q; |
| 80 | rho_1 = rho; |
| 81 | rho = r.dot(r); |
| 82 | p = (rho/rho_1) * p + r; |
| 83 | } |
| 84 | |
| 85 | l = C.transpose() * e; // l is the i-th row of CINV |
| 86 | // FIXME add a generic "prune/filter" expression for both dense and sparse object to sparse |
| 87 | for (Index j=0; j<l.size(); ++j) |
| 88 | if (l[j]<1e-15) |
| 89 | tripletList.push_back(T(i,j,l(j))); |
| 90 | |
| 91 | |
| 92 | d[i] = 0.0; |
| 93 | } |
| 94 | CINV.setFromTriplets(tripletList.begin(), tripletList.end()); |
| 95 | } |
| 96 | |
| 97 | |
| 98 | |
| 99 | /** \ingroup IterativeSolvers_Module |
| 100 | * Constrained conjugate gradient |
| 101 | * |
| 102 | * Computes the minimum of \f$ 1/2((Ax).x) - bx \f$ under the contraint \f$ Cx \le f \f$ |
| 103 | */ |
| 104 | template<typename TMatrix, typename CMatrix, |
| 105 | typename VectorX, typename VectorB, typename VectorF> |
| 106 | void constrained_cg(const TMatrix& A, const CMatrix& C, VectorX& x, |
| 107 | const VectorB& b, const VectorF& f, IterationController &iter) |
| 108 | { |
| 109 | using std::sqrt; |
| 110 | typedef typename TMatrix::Scalar Scalar; |
| 111 | typedef typename TMatrix::Index Index; |
| 112 | typedef Matrix<Scalar,Dynamic,1> TmpVec; |
| 113 | |
| 114 | Scalar rho = 1.0, rho_1, lambda, gamma; |
| 115 | Index xSize = x.size(); |
| 116 | TmpVec p(xSize), q(xSize), q2(xSize), |
| 117 | r(xSize), old_z(xSize), z(xSize), |
| 118 | memox(xSize); |
| 119 | std::vector<bool> satured(C.rows()); |
| 120 | p.setZero(); |
| 121 | iter.setRhsNorm(sqrt(b.dot(b))); // gael vect_sp(PS, b, b) |
| 122 | if (iter.rhsNorm() == 0.0) iter.setRhsNorm(1.0); |
| 123 | |
| 124 | SparseMatrix<Scalar,RowMajor> CINV(C.rows(), C.cols()); |
| 125 | pseudo_inverse(C, CINV); |
| 126 | |
| 127 | while(true) |
| 128 | { |
| 129 | // computation of residual |
| 130 | old_z = z; |
| 131 | memox = x; |
| 132 | r = b; |
| 133 | r += A * -x; |
| 134 | z = r; |
| 135 | bool transition = false; |
| 136 | for (Index i = 0; i < C.rows(); ++i) |
| 137 | { |
| 138 | Scalar al = C.row(i).dot(x) - f.coeff(i); |
| 139 | if (al >= -1.0E-15) |
| 140 | { |
| 141 | if (!satured[i]) |
| 142 | { |
| 143 | satured[i] = true; |
| 144 | transition = true; |
| 145 | } |
| 146 | Scalar bb = CINV.row(i).dot(z); |
| 147 | if (bb > 0.0) |
| 148 | // FIXME: we should allow that: z += -bb * C.row(i); |
| 149 | for (typename CMatrix::InnerIterator it(C,i); it; ++it) |
| 150 | z.coeffRef(it.index()) -= bb*it.value(); |
| 151 | } |
| 152 | else |
| 153 | satured[i] = false; |
| 154 | } |
| 155 | |
| 156 | // descent direction |
| 157 | rho_1 = rho; |
| 158 | rho = r.dot(z); |
| 159 | |
| 160 | if (iter.finished(rho)) break; |
| 161 | |
| 162 | if (iter.noiseLevel() > 0 && transition) std::cerr << "CCG: transition\n"; |
| 163 | if (transition || iter.first()) gamma = 0.0; |
| 164 | else gamma = (std::max)(0.0, (rho - old_z.dot(z)) / rho_1); |
| 165 | p = z + gamma*p; |
| 166 | |
| 167 | ++iter; |
| 168 | // one dimensionnal optimization |
| 169 | q = A * p; |
| 170 | lambda = rho / q.dot(p); |
| 171 | for (Index i = 0; i < C.rows(); ++i) |
| 172 | { |
| 173 | if (!satured[i]) |
| 174 | { |
| 175 | Scalar bb = C.row(i).dot(p) - f[i]; |
| 176 | if (bb > 0.0) |
| 177 | lambda = (std::min)(lambda, (f.coeff(i)-C.row(i).dot(x)) / bb); |
| 178 | } |
| 179 | } |
| 180 | x += lambda * p; |
| 181 | memox -= x; |
| 182 | } |
| 183 | } |
| 184 | |
| 185 | } // end namespace internal |
| 186 | |
| 187 | } // end namespace Eigen |
| 188 | |
| 189 | #endif // EIGEN_CONSTRAINEDCG_H |