#include <DiagonalLowRankSolver.h>

Inheritance diagram for dp3::ddecal::DiagonalLowRankSolver:

Public Member Functions
	DiagonalLowRankSolver (size_t n_low_rank_approximation_iterations, size_t n_power_iterations)

size_t	NSolutionPolarizations () const override

void	SetNLowRankApproximationIterations (size_t n_iterations)

void	SetNPowerIterations (size_t n_power_iterations)

SolveResult	Solve (const FullSolveData &data, std::vector< std::vector< DComplex >> &solutions, double time) override

Public Member Functions inherited from dp3::ddecal::SolverBase
	SolverBase ()

virtual	~SolverBase ()=default

void	AddConstraint (std::unique_ptr< Constraint > constraint)

virtual std::vector< SolverBase * >	ConstraintSolvers ()

const std::vector< std::unique_ptr< Constraint > > &	GetConstraints ()

LLSSolverType	GetLLSSolverType () const

bool	GetPhaseOnly () const

void	GetTimings (std::ostream &os, double duration) const

virtual void	Initialize (size_t n_antennas, const std::vector< size_t > &n_solutions_per_direction, size_t n_channel_blocks)

void	SetDdConstraintWeights (const std::vector< std::vector< double >> &weights)

void	SetLLSSolverType (LLSSolverType solver_type)

void	SetPhaseOnly (bool phase_only)

virtual SolveResult	Solve (const UniSolveData &data, std::vector< std::vector< DComplex >> &solutions, double time)

virtual bool	SupportsDdSolutionIntervals () const

size_t	GetMaxIterations () const

void	SetMaxIterations (size_t max_iterations)

size_t	GetMinIterations () const

void	SetMinIterations (size_t min_iterations)

void	SetAccuracy (double accuracy)

double	GetAccuracy () const

void	SetConstraintAccuracy (double constraint_accuracy)

double	GetConstraintAccuracy () const

void	SetStepSize (double step_size)

double	GetStepSize () const

void	SetDetectStalling (bool detect_stalling, double step_diff_sigma)

bool	GetDetectStalling () const

Additional Inherited Members
Public Types inherited from dp3::ddecal::SolverBase
typedef std::complex< float >	Complex

typedef std::complex< double >	DComplex

Protected Member Functions inherited from dp3::ddecal::SolverBase
bool	ApplyConstraints (size_t iteration, double time, bool has_previously_converged, SolutionSpan &next_solutions) const

bool	ApplyConstraints (size_t iteration, double time, bool has_previously_converged, SolutionTensor &next_solutions) const

bool	AssignSolutions (std::vector< std::vector< DComplex >> &solutions, SolutionSpan &new_solutions, bool use_constraint_accuracy, double &avg_abs_diff, std::vector< double > &step_magnitudes) const

bool	AssignSolutions (std::vector< std::vector< DComplex >> &solutions, SolutionTensor &new_solutions, bool use_constraint_accuracy, double &avg_abs_diff, std::vector< double > &step_magnitudes) const

std::unique_ptr< LLSSolver >	CreateLLSSolver (size_t m, size_t n, size_t nrhs) const

bool	DetectStall (size_t iteration, const std::vector< double > &step_magnitudes)

std::unique_ptr< aocommon::RecursiveFor >	MakeOptionalRecursiveFor () const

SolveResult	MakeResult (size_t iteration, bool has_converged, bool constraints_satisfied) const

size_t	NAntennas () const

size_t	NChannelBlocks () const

size_t	NDirections () const

size_t	NSubSolutions () const

size_t	NSubThreads () const

size_t	NVisibilities () const

void	PrepareConstraints ()

bool	ReachedStoppingCriterion (size_t iteration, bool has_converged, bool constraints_satisfied, const std::vector< double > &step_magnitudes)

void	Step (const std::vector< std::vector< DComplex >> &solutions, SolutionTensor &next_solutions) const

Static Protected Member Functions inherited from dp3::ddecal::SolverBase
static void	MakeSolutionsFinite1Pol (std::vector< std::vector< DComplex >> &solutions)

static void	MakeSolutionsFinite2Pol (std::vector< std::vector< DComplex >> &solutions)

static void	MakeSolutionsFinite4Pol (std::vector< std::vector< DComplex >> &solutions)

Detailed Description

Experimental solver that make use of the property that the antenna x antenna visibility matrix is a low-rank (rank 1 to be precise) matrix. A normal low-rank approximation can be solved exactly and without iterations, but because there are weights and missing values in the matrix, an iterative procedure is required, and the result is not exact.

When the step size is set to 1, this solver makes an immediate step. Because a solving iteration uses all data for one direction at once, this may be a good approach (in contrast to the iterative diagonal solver). To help convergence in this case, the directions are solved in order of largest power first.

Constructor & Destructor Documentation

◆ DiagonalLowRankSolver()

dp3::ddecal::DiagonalLowRankSolver::DiagonalLowRankSolver	(	size_t	n_low_rank_approximation_iterations,
		size_t	n_power_iterations
	)

inline

Member Function Documentation

◆ NSolutionPolarizations()

size_t dp3::ddecal::DiagonalLowRankSolver::NSolutionPolarizations ( ) const

inlineoverridevirtual

Returns: The number of polarizations in the solution.

Implements dp3::ddecal::SolverBase.

◆ SetNLowRankApproximationIterations()

void dp3::ddecal::DiagonalLowRankSolver::SetNLowRankApproximationIterations ( size_t n_iterations )

inline

Number of full low-rank approximation iterations.

◆ SetNPowerIterations()

void dp3::ddecal::DiagonalLowRankSolver::SetNPowerIterations ( size_t n_power_iterations )

inline

Number of power iterations performed to calculate the eigen value. See also DominantEigenPair().

◆ Solve()

SolveResult dp3::ddecal::DiagonalLowRankSolver::Solve	(	const FullSolveData &	data,
		std::vector< std::vector< DComplex >> &	solutions,
		double	time
	)

overridevirtual

Solves multi-directional Jones matrices. Takes the (single) measured data and the (multi-directional) model data, and solves the optimization problem that minimizes the norm of the differences.

Parameters

data	Buffer with weighted data and model data.
solutions	The per-channel and per-antenna solutions. solutions[ch] is a pointer for channelblock ch to antenna x directions x pol solutions.

Reimplemented from dp3::ddecal::SolverBase.

The documentation for this class was generated from the following file:

DiagonalLowRankSolver.h

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ DiagonalLowRankSolver()

Member Function Documentation

◆ NSolutionPolarizations()

◆ SetNLowRankApproximationIterations()

◆ SetNPowerIterations()

◆ Solve()