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 (at) inria.fr> 5 // 6 // This Source Code Form is subject to the terms of the Mozilla 7 // Public License v. 2.0. If a copy of the MPL was not distributed 8 // with this file, You can obtain one at http://mozilla.org/MPL/2.0/. 9 10 #ifndef EIGEN_LLT_H 11 #define EIGEN_LLT_H 12 13 namespace Eigen { 14 15 namespace internal{ 16 template<typename MatrixType, int UpLo> struct LLT_Traits; 17 } 18 19 /** \ingroup Cholesky_Module 20 * 21 * \class LLT 22 * 23 * \brief Standard Cholesky decomposition (LL^T) of a matrix and associated features 24 * 25 * \param MatrixType the type of the matrix of which we are computing the LL^T Cholesky decomposition 26 * \param UpLo the triangular part that will be used for the decompositon: Lower (default) or Upper. 27 * The other triangular part won't be read. 28 * 29 * This class performs a LL^T Cholesky decomposition of a symmetric, positive definite 30 * matrix A such that A = LL^* = U^*U, where L is lower triangular. 31 * 32 * While the Cholesky decomposition is particularly useful to solve selfadjoint problems like D^*D x = b, 33 * for that purpose, we recommend the Cholesky decomposition without square root which is more stable 34 * and even faster. Nevertheless, this standard Cholesky decomposition remains useful in many other 35 * situations like generalised eigen problems with hermitian matrices. 36 * 37 * Remember that Cholesky decompositions are not rank-revealing. This LLT decomposition is only stable on positive definite matrices, 38 * use LDLT instead for the semidefinite case. Also, do not use a Cholesky decomposition to determine whether a system of equations 39 * has a solution. 40 * 41 * Example: \include LLT_example.cpp 42 * Output: \verbinclude LLT_example.out 43 * 44 * \sa MatrixBase::llt(), class LDLT 45 */ 46 /* HEY THIS DOX IS DISABLED BECAUSE THERE's A BUG EITHER HERE OR IN LDLT ABOUT THAT (OR BOTH) 47 * Note that during the decomposition, only the upper triangular part of A is considered. Therefore, 48 * the strict lower part does not have to store correct values. 49 */ 50 template<typename _MatrixType, int _UpLo> class LLT 51 { 52 public: 53 typedef _MatrixType MatrixType; 54 enum { 55 RowsAtCompileTime = MatrixType::RowsAtCompileTime, 56 ColsAtCompileTime = MatrixType::ColsAtCompileTime, 57 Options = MatrixType::Options, 58 MaxColsAtCompileTime = MatrixType::MaxColsAtCompileTime 59 }; 60 typedef typename MatrixType::Scalar Scalar; 61 typedef typename NumTraits<typename MatrixType::Scalar>::Real RealScalar; 62 typedef typename MatrixType::Index Index; 63 64 enum { 65 PacketSize = internal::packet_traits<Scalar>::size, 66 AlignmentMask = int(PacketSize)-1, 67 UpLo = _UpLo 68 }; 69 70 typedef internal::LLT_Traits<MatrixType,UpLo> Traits; 71 72 /** 73 * \brief Default Constructor. 74 * 75 * The default constructor is useful in cases in which the user intends to 76 * perform decompositions via LLT::compute(const MatrixType&). 77 */ 78 LLT() : m_matrix(), m_isInitialized(false) {} 79 80 /** \brief Default Constructor with memory preallocation 81 * 82 * Like the default constructor but with preallocation of the internal data 83 * according to the specified problem \a size. 84 * \sa LLT() 85 */ 86 LLT(Index size) : m_matrix(size, size), 87 m_isInitialized(false) {} 88 89 LLT(const MatrixType& matrix) 90 : m_matrix(matrix.rows(), matrix.cols()), 91 m_isInitialized(false) 92 { 93 compute(matrix); 94 } 95 96 /** \returns a view of the upper triangular matrix U */ 97 inline typename Traits::MatrixU matrixU() const 98 { 99 eigen_assert(m_isInitialized && "LLT is not initialized."); 100 return Traits::getU(m_matrix); 101 } 102 103 /** \returns a view of the lower triangular matrix L */ 104 inline typename Traits::MatrixL matrixL() const 105 { 106 eigen_assert(m_isInitialized && "LLT is not initialized."); 107 return Traits::getL(m_matrix); 108 } 109 110 /** \returns the solution x of \f$ A x = b \f$ using the current decomposition of A. 111 * 112 * Since this LLT class assumes anyway that the matrix A is invertible, the solution 113 * theoretically exists and is unique regardless of b. 114 * 115 * Example: \include LLT_solve.cpp 116 * Output: \verbinclude LLT_solve.out 117 * 118 * \sa solveInPlace(), MatrixBase::llt() 119 */ 120 template<typename Rhs> 121 inline const internal::solve_retval<LLT, Rhs> 122 solve(const MatrixBase<Rhs>& b) const 123 { 124 eigen_assert(m_isInitialized && "LLT is not initialized."); 125 eigen_assert(m_matrix.rows()==b.rows() 126 && "LLT::solve(): invalid number of rows of the right hand side matrix b"); 127 return internal::solve_retval<LLT, Rhs>(*this, b.derived()); 128 } 129 130 #ifdef EIGEN2_SUPPORT 131 template<typename OtherDerived, typename ResultType> 132 bool solve(const MatrixBase<OtherDerived>& b, ResultType *result) const 133 { 134 *result = this->solve(b); 135 return true; 136 } 137 138 bool isPositiveDefinite() const { return true; } 139 #endif 140 141 template<typename Derived> 142 void solveInPlace(MatrixBase<Derived> &bAndX) const; 143 144 LLT& compute(const MatrixType& matrix); 145 146 /** \returns the LLT decomposition matrix 147 * 148 * TODO: document the storage layout 149 */ 150 inline const MatrixType& matrixLLT() const 151 { 152 eigen_assert(m_isInitialized && "LLT is not initialized."); 153 return m_matrix; 154 } 155 156 MatrixType reconstructedMatrix() const; 157 158 159 /** \brief Reports whether previous computation was successful. 160 * 161 * \returns \c Success if computation was succesful, 162 * \c NumericalIssue if the matrix.appears to be negative. 163 */ 164 ComputationInfo info() const 165 { 166 eigen_assert(m_isInitialized && "LLT is not initialized."); 167 return m_info; 168 } 169 170 inline Index rows() const { return m_matrix.rows(); } 171 inline Index cols() const { return m_matrix.cols(); } 172 173 template<typename VectorType> 174 LLT rankUpdate(const VectorType& vec, const RealScalar& sigma = 1); 175 176 protected: 177 /** \internal 178 * Used to compute and store L 179 * The strict upper part is not used and even not initialized. 180 */ 181 MatrixType m_matrix; 182 bool m_isInitialized; 183 ComputationInfo m_info; 184 }; 185 186 namespace internal { 187 188 template<typename Scalar, int UpLo> struct llt_inplace; 189 190 template<typename MatrixType, typename VectorType> 191 static typename MatrixType::Index llt_rank_update_lower(MatrixType& mat, const VectorType& vec, const typename MatrixType::RealScalar& sigma) 192 { 193 typedef typename MatrixType::Scalar Scalar; 194 typedef typename MatrixType::RealScalar RealScalar; 195 typedef typename MatrixType::Index Index; 196 typedef typename MatrixType::ColXpr ColXpr; 197 typedef typename internal::remove_all<ColXpr>::type ColXprCleaned; 198 typedef typename ColXprCleaned::SegmentReturnType ColXprSegment; 199 typedef Matrix<Scalar,Dynamic,1> TempVectorType; 200 typedef typename TempVectorType::SegmentReturnType TempVecSegment; 201 202 int n = mat.cols(); 203 eigen_assert(mat.rows()==n && vec.size()==n); 204 205 TempVectorType temp; 206 207 if(sigma>0) 208 { 209 // This version is based on Givens rotations. 210 // It is faster than the other one below, but only works for updates, 211 // i.e., for sigma > 0 212 temp = sqrt(sigma) * vec; 213 214 for(int i=0; i<n; ++i) 215 { 216 JacobiRotation<Scalar> g; 217 g.makeGivens(mat(i,i), -temp(i), &mat(i,i)); 218 219 int rs = n-i-1; 220 if(rs>0) 221 { 222 ColXprSegment x(mat.col(i).tail(rs)); 223 TempVecSegment y(temp.tail(rs)); 224 apply_rotation_in_the_plane(x, y, g); 225 } 226 } 227 } 228 else 229 { 230 temp = vec; 231 RealScalar beta = 1; 232 for(int j=0; j<n; ++j) 233 { 234 RealScalar Ljj = real(mat.coeff(j,j)); 235 RealScalar dj = abs2(Ljj); 236 Scalar wj = temp.coeff(j); 237 RealScalar swj2 = sigma*abs2(wj); 238 RealScalar gamma = dj*beta + swj2; 239 240 RealScalar x = dj + swj2/beta; 241 if (x<=RealScalar(0)) 242 return j; 243 RealScalar nLjj = sqrt(x); 244 mat.coeffRef(j,j) = nLjj; 245 beta += swj2/dj; 246 247 // Update the terms of L 248 Index rs = n-j-1; 249 if(rs) 250 { 251 temp.tail(rs) -= (wj/Ljj) * mat.col(j).tail(rs); 252 if(gamma != 0) 253 mat.col(j).tail(rs) = (nLjj/Ljj) * mat.col(j).tail(rs) + (nLjj * sigma*conj(wj)/gamma)*temp.tail(rs); 254 } 255 } 256 } 257 return -1; 258 } 259 260 template<typename Scalar> struct llt_inplace<Scalar, Lower> 261 { 262 typedef typename NumTraits<Scalar>::Real RealScalar; 263 template<typename MatrixType> 264 static typename MatrixType::Index unblocked(MatrixType& mat) 265 { 266 typedef typename MatrixType::Index Index; 267 268 eigen_assert(mat.rows()==mat.cols()); 269 const Index size = mat.rows(); 270 for(Index k = 0; k < size; ++k) 271 { 272 Index rs = size-k-1; // remaining size 273 274 Block<MatrixType,Dynamic,1> A21(mat,k+1,k,rs,1); 275 Block<MatrixType,1,Dynamic> A10(mat,k,0,1,k); 276 Block<MatrixType,Dynamic,Dynamic> A20(mat,k+1,0,rs,k); 277 278 RealScalar x = real(mat.coeff(k,k)); 279 if (k>0) x -= A10.squaredNorm(); 280 if (x<=RealScalar(0)) 281 return k; 282 mat.coeffRef(k,k) = x = sqrt(x); 283 if (k>0 && rs>0) A21.noalias() -= A20 * A10.adjoint(); 284 if (rs>0) A21 *= RealScalar(1)/x; 285 } 286 return -1; 287 } 288 289 template<typename MatrixType> 290 static typename MatrixType::Index blocked(MatrixType& m) 291 { 292 typedef typename MatrixType::Index Index; 293 eigen_assert(m.rows()==m.cols()); 294 Index size = m.rows(); 295 if(size<32) 296 return unblocked(m); 297 298 Index blockSize = size/8; 299 blockSize = (blockSize/16)*16; 300 blockSize = (std::min)((std::max)(blockSize,Index(8)), Index(128)); 301 302 for (Index k=0; k<size; k+=blockSize) 303 { 304 // partition the matrix: 305 // A00 | - | - 306 // lu = A10 | A11 | - 307 // A20 | A21 | A22 308 Index bs = (std::min)(blockSize, size-k); 309 Index rs = size - k - bs; 310 Block<MatrixType,Dynamic,Dynamic> A11(m,k, k, bs,bs); 311 Block<MatrixType,Dynamic,Dynamic> A21(m,k+bs,k, rs,bs); 312 Block<MatrixType,Dynamic,Dynamic> A22(m,k+bs,k+bs,rs,rs); 313 314 Index ret; 315 if((ret=unblocked(A11))>=0) return k+ret; 316 if(rs>0) A11.adjoint().template triangularView<Upper>().template solveInPlace<OnTheRight>(A21); 317 if(rs>0) A22.template selfadjointView<Lower>().rankUpdate(A21,-1); // bottleneck 318 } 319 return -1; 320 } 321 322 template<typename MatrixType, typename VectorType> 323 static typename MatrixType::Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma) 324 { 325 return Eigen::internal::llt_rank_update_lower(mat, vec, sigma); 326 } 327 }; 328 329 template<typename Scalar> struct llt_inplace<Scalar, Upper> 330 { 331 typedef typename NumTraits<Scalar>::Real RealScalar; 332 333 template<typename MatrixType> 334 static EIGEN_STRONG_INLINE typename MatrixType::Index unblocked(MatrixType& mat) 335 { 336 Transpose<MatrixType> matt(mat); 337 return llt_inplace<Scalar, Lower>::unblocked(matt); 338 } 339 template<typename MatrixType> 340 static EIGEN_STRONG_INLINE typename MatrixType::Index blocked(MatrixType& mat) 341 { 342 Transpose<MatrixType> matt(mat); 343 return llt_inplace<Scalar, Lower>::blocked(matt); 344 } 345 template<typename MatrixType, typename VectorType> 346 static typename MatrixType::Index rankUpdate(MatrixType& mat, const VectorType& vec, const RealScalar& sigma) 347 { 348 Transpose<MatrixType> matt(mat); 349 return llt_inplace<Scalar, Lower>::rankUpdate(matt, vec.conjugate(), sigma); 350 } 351 }; 352 353 template<typename MatrixType> struct LLT_Traits<MatrixType,Lower> 354 { 355 typedef const TriangularView<const MatrixType, Lower> MatrixL; 356 typedef const TriangularView<const typename MatrixType::AdjointReturnType, Upper> MatrixU; 357 static inline MatrixL getL(const MatrixType& m) { return m; } 358 static inline MatrixU getU(const MatrixType& m) { return m.adjoint(); } 359 static bool inplace_decomposition(MatrixType& m) 360 { return llt_inplace<typename MatrixType::Scalar, Lower>::blocked(m)==-1; } 361 }; 362 363 template<typename MatrixType> struct LLT_Traits<MatrixType,Upper> 364 { 365 typedef const TriangularView<const typename MatrixType::AdjointReturnType, Lower> MatrixL; 366 typedef const TriangularView<const MatrixType, Upper> MatrixU; 367 static inline MatrixL getL(const MatrixType& m) { return m.adjoint(); } 368 static inline MatrixU getU(const MatrixType& m) { return m; } 369 static bool inplace_decomposition(MatrixType& m) 370 { return llt_inplace<typename MatrixType::Scalar, Upper>::blocked(m)==-1; } 371 }; 372 373 } // end namespace internal 374 375 /** Computes / recomputes the Cholesky decomposition A = LL^* = U^*U of \a matrix 376 * 377 * \returns a reference to *this 378 * 379 * Example: \include TutorialLinAlgComputeTwice.cpp 380 * Output: \verbinclude TutorialLinAlgComputeTwice.out 381 */ 382 template<typename MatrixType, int _UpLo> 383 LLT<MatrixType,_UpLo>& LLT<MatrixType,_UpLo>::compute(const MatrixType& a) 384 { 385 eigen_assert(a.rows()==a.cols()); 386 const Index size = a.rows(); 387 m_matrix.resize(size, size); 388 m_matrix = a; 389 390 m_isInitialized = true; 391 bool ok = Traits::inplace_decomposition(m_matrix); 392 m_info = ok ? Success : NumericalIssue; 393 394 return *this; 395 } 396 397 /** Performs a rank one update (or dowdate) of the current decomposition. 398 * If A = LL^* before the rank one update, 399 * then after it we have LL^* = A + sigma * v v^* where \a v must be a vector 400 * of same dimension. 401 */ 402 template<typename _MatrixType, int _UpLo> 403 template<typename VectorType> 404 LLT<_MatrixType,_UpLo> LLT<_MatrixType,_UpLo>::rankUpdate(const VectorType& v, const RealScalar& sigma) 405 { 406 EIGEN_STATIC_ASSERT_VECTOR_ONLY(VectorType); 407 eigen_assert(v.size()==m_matrix.cols()); 408 eigen_assert(m_isInitialized); 409 if(internal::llt_inplace<typename MatrixType::Scalar, UpLo>::rankUpdate(m_matrix,v,sigma)>=0) 410 m_info = NumericalIssue; 411 else 412 m_info = Success; 413 414 return *this; 415 } 416 417 namespace internal { 418 template<typename _MatrixType, int UpLo, typename Rhs> 419 struct solve_retval<LLT<_MatrixType, UpLo>, Rhs> 420 : solve_retval_base<LLT<_MatrixType, UpLo>, Rhs> 421 { 422 typedef LLT<_MatrixType,UpLo> LLTType; 423 EIGEN_MAKE_SOLVE_HELPERS(LLTType,Rhs) 424 425 template<typename Dest> void evalTo(Dest& dst) const 426 { 427 dst = rhs(); 428 dec().solveInPlace(dst); 429 } 430 }; 431 } 432 433 /** \internal use x = llt_object.solve(x); 434 * 435 * This is the \em in-place version of solve(). 436 * 437 * \param bAndX represents both the right-hand side matrix b and result x. 438 * 439 * \returns true always! If you need to check for existence of solutions, use another decomposition like LU, QR, or SVD. 440 * 441 * This version avoids a copy when the right hand side matrix b is not 442 * needed anymore. 443 * 444 * \sa LLT::solve(), MatrixBase::llt() 445 */ 446 template<typename MatrixType, int _UpLo> 447 template<typename Derived> 448 void LLT<MatrixType,_UpLo>::solveInPlace(MatrixBase<Derived> &bAndX) const 449 { 450 eigen_assert(m_isInitialized && "LLT is not initialized."); 451 eigen_assert(m_matrix.rows()==bAndX.rows()); 452 matrixL().solveInPlace(bAndX); 453 matrixU().solveInPlace(bAndX); 454 } 455 456 /** \returns the matrix represented by the decomposition, 457 * i.e., it returns the product: L L^*. 458 * This function is provided for debug purpose. */ 459 template<typename MatrixType, int _UpLo> 460 MatrixType LLT<MatrixType,_UpLo>::reconstructedMatrix() const 461 { 462 eigen_assert(m_isInitialized && "LLT is not initialized."); 463 return matrixL() * matrixL().adjoint().toDenseMatrix(); 464 } 465 466 /** \cholesky_module 467 * \returns the LLT decomposition of \c *this 468 */ 469 template<typename Derived> 470 inline const LLT<typename MatrixBase<Derived>::PlainObject> 471 MatrixBase<Derived>::llt() const 472 { 473 return LLT<PlainObject>(derived()); 474 } 475 476 /** \cholesky_module 477 * \returns the LLT decomposition of \c *this 478 */ 479 template<typename MatrixType, unsigned int UpLo> 480 inline const LLT<typename SelfAdjointView<MatrixType, UpLo>::PlainObject, UpLo> 481 SelfAdjointView<MatrixType, UpLo>::llt() const 482 { 483 return LLT<PlainObject,UpLo>(m_matrix); 484 } 485 486 } // end namespace Eigen 487 488 #endif // EIGEN_LLT_H 489
