#include <LOBPCGSolver.h>

Public Member Functions
void	compute (int maxit=10, Scalar tol_div_n=1e-7)

Vector	eigenvalues ()

Matrix	eigenvectors ()

int	info ()

	LOBPCGSolver (const SparseMatrix &A, const SparseMatrix X)

Matrix	residuals ()

void	setB (const SparseMatrix &B)

void	setConstraints (const SparseMatrix &Y)

void	setPreconditioner (const SparseMatrix &preconditioner)

Public Attributes
Vector	m_evalues

Matrix	m_evectors

int	m_info

SparseMatrix	m_residuals

Private Types
typedef std::complex< Scalar >	Complex

typedef Eigen::Matrix< Complex, Eigen::Dynamic, Eigen::Dynamic >	ComplexMatrix

typedef Eigen::Matrix< Complex, Eigen::Dynamic, 1 >	ComplexVector

typedef Eigen::Matrix< Scalar, Eigen::Dynamic, Eigen::Dynamic >	Matrix

typedef Eigen::SparseMatrix< Complex >	SparseComplexMatrix

typedef Eigen::SparseMatrix< Scalar >	SparseMatrix

typedef Eigen::Matrix< Scalar, Eigen::Dynamic, 1 >	Vector

Private Member Functions
void	applyConstraintsInPlace (SparseMatrix &X, SparseMatrix &Y, SparseMatrix &B)

int	checkConvergence_getBlocksize (SparseMatrix &m_residuals, Scalar tolerance_L2, std::vector< int > &columnsToDelete)

int	orthogonalizeInPlace (SparseMatrix &M, SparseMatrix &B, SparseMatrix &true_BM, bool has_true_BM=false)

void	removeColumns (SparseMatrix &matrix, std::vector< int > &colToRemove)

void	sort_epairs (Vector &evalues, Matrix &evectors, int SelectionRule)

Matrix	stack_4_matricies (Matrix A, Matrix B, Matrix C, Matrix D)

Matrix	stack_9_matricies (Matrix A, Matrix B, Matrix C, Matrix D, Matrix E, Matrix F, Matrix G, Matrix H, Matrix I)

Private Attributes
SparseMatrix	A

bool	flag_with_B

bool	flag_with_constraints

bool	flag_with_preconditioner

SparseMatrix	m_B

const int	m_n

const int	m_nev

SparseMatrix	m_preconditioner

SparseMatrix	m_Y

SparseMatrix	X

Detailed Description

template<typename Scalar = long double>
class Spectra::LOBPCGSolver< Scalar >

*** METHOD The class represent the LOBPCG algorithm, which was invented by Andrew Knyazev Theoretical background of the procedure can be found in the articles below

Knyazev, A.V., 2001. Toward the optimal preconditioned eigensolver : Locally optimal block preconditioned conjugate gradient method.SIAM journal on scientific computing, 23(2), pp.517 - 541.
Knyazev, A.V., Argentati, M.E., Lashuk, I. and Ovtchinnikov, E.E., 2007. Block locally optimal preconditioned eigenvalue xolvers(BLOPEX) in HYPRE and PETSc.SIAM Journal on Scientific Computing, 29(5), pp.2224 - 2239.

*** CONDITIONS OF USE Locally Optimal Block Preconditioned Conjugate Gradient(LOBPCG) is a method for finding the M smallest eigenvalues and eigenvectors of a large symmetric positive definite generalized eigenvalue problem $Ax=\lambda Bx,$ where $A_{NxN}$ is a symmetric matrix, $B$ is symmetric and positive - definite. $A and B$ are also assumed large and sparse $\textit{M}$ should be $\<< textit{N}$ (at least $\textit{5M} < \textit{N}$ )

*** ARGUMENTS Eigen::SparseMatrix<long double> A; // N*N - Ax = lambda*Bx, lrage and sparse Eigen::SparseMatrix<long double> X; // N*M - initial approximations to eigenvectors (random in general case) Spectra::LOBPCGSolver<long double> solver(A, X); *Eigen::SparseMatrix<long double> B; // N*N - Ax = lambda*Bx, sparse, positive definite solver.setConstraints(B); *Eigen::SparseMatrix<long double> Y; // N*K - constraints, already found eigenvectors solver.setB(B); *Eigen::SparseMatrix<long double> T; // N*N - preconditioner ~ A^-1 solver.setPreconditioner(T);

*** OUTCOMES solver.solve(); // compute eigenpairs // void solver.info(); // state of converjance // int solver.residuals(); // get residuals to evaluate biases // Eigen::Matrix<Scalar, Eigen::Dynamic, Eigen::Dynamic> solver.eigenvalues(); // get eigenvalues // Eigen::Matrix<Scalar, Eigen::Dynamic, Eigen::Dynamic> solver.eigenvectors(); // get eigenvectors // Eigen::Matrix<Scalar, Eigen::Dynamic, 1>

*** EXAMPLE

#include <Spectra/contrib/SymSparseEigsSolverLOBPCG.h>
    // random A
    Matrix a;
    a = (Matrix::Random(10, 10).array() > 0.6).cast<long double>() * Matrix::Random(10, 10).array() * 5;
    a = Matrix((a).triangularView<Eigen::Lower>()) + Matrix((a).triangularView<Eigen::Lower>()).transpose();
    for (int i = 0; i < 10; i++)
        a(i, i) = i + 0.5;
    std::cout << a << "\n";
    Eigen::SparseMatrix<long double> A(a.sparseView());
    // random X
    Eigen::Matrix<long double, 10, 2> x;
    x = Matrix::Random(10, 2).array();
    Eigen::SparseMatrix<long double> X(x.sparseView());
    // solve Ax = lambda*x
    Spectra::LOBPCGSolver<long double> solver(A, X);
    solver.compute(10, 1e-4); // 10 iterations, L2_tolerance = 1e-4*N
std::cout << "info\n" << solver.info() << std::endl;
std::cout << "eigenvalues\n" << solver.eigenvalues() << std::endl;
std::cout << "eigenvectors\n" << solver.eigenvectors() << std::endl;
std::cout << "residuals\n" << solver.residuals() << std::endl;

Definition at line 83 of file LOBPCGSolver.h.