A class that defines an optimization algorithm and its settings. This class defines the settings for an optimization algorithm, including whether to use nonlinear conjugate gradient (NCG), whether NCG has gotten stuck and cannot further minimize, and whether to use R-pi adaptive method for regularization energy. These settings can be accessed and modified using the public member variables of the class. More...

#include <Model.hpp>

Public Member Functions
	OptimizationAlgorithm ()
	Constructor for the OptimizationAlgorithm class.

	OptimizationAlgorithm (const int trialIterationInterval, const double c1, const double c2, const double stepThreshold)
	Constructor for the OptimizationAlgorithm class.

void	reset_NCG_Rpi ()
	Reset NCG and Rpi to default.

bool	disable_NCG_if_stuck_consecutively ()
	Disables the use of NCG if it has been stuck consecutively over a threshold.

Public Attributes
bool	usingNCG = false
	Flag indicating whether to use the nonlinear conjugate gradient optimizer.

bool	isNCGstuck = false
	Flag indicating whether the NCG optimizer has gotten stuck and cannot further minimize.

bool	usingRpi = false
	Flag indicating whether to use the R-pi adaptive method for regularization energy.

bool	isCriteriaSatisfied = false
	Flag indicating whether the optimization criteria have been met.

double	trialStepSize = 0.0
	The step size used in the optimization algorithm.

int	trialIterationInterval = 300
	The number of iterations between trial step size adjustments.

double	c1 = 1e-6
	The c1 parameter in the NCG Wolfe condition.

double	c2 = 0.001
	The c2 parameter in the NCG Wolfe condition.

double	stepThreshold = 1e-15
	The lower bound for the step size in NCG.

int	nConsecutiveNcgStuck = 0
	The number of consecutive times the NCG optimizer has gotten stuck.

int	nConsecutiveNcgStuckThreshold = 3
	The threshold of consecutive NCG stuck iterations before disabling it.

double	energyDiffThreshold = 1e-5
	The energy difference threshold for convergence checking.

double	forceDiffThreshold = 1e-4
	The mean force difference threshold for convergence checking.

Detailed Description

A class that defines an optimization algorithm and its settings. This class defines the settings for an optimization algorithm, including whether to use nonlinear conjugate gradient (NCG), whether NCG has gotten stuck and cannot further minimize, and whether to use R-pi adaptive method for regularization energy. These settings can be accessed and modified using the public member variables of the class.

Constructor & Destructor Documentation

◆ OptimizationAlgorithm() [1/2]

OptimizationAlgorithm::OptimizationAlgorithm ( )

Constructor for the OptimizationAlgorithm class.

Initializes an instance of the OptimizationAlgorithm class with default settings. Sets usingNCG to true, isNCGstuck to false, and usingRpi to true.

◆ OptimizationAlgorithm() [2/2]

OptimizationAlgorithm::OptimizationAlgorithm	(	const int	trialIterationInterval,
		const double	c1,
		const double	c2,
		const double	stepThreshold
	)

Constructor for the OptimizationAlgorithm class.

Initializes an instance of the OptimizationAlgorithm class with default settings. Customize iteration interval, c1, c2, and step threshold. Sets usingNCG to true, isNCGstuck to false, and usingRpi to true.

Member Function Documentation

◆ disable_NCG_if_stuck_consecutively()

bool OptimizationAlgorithm::disable_NCG_if_stuck_consecutively ( )

Disables the use of NCG if it has been stuck consecutively over a threshold.

If the NCG optimizer has been stuck for too many consecutive iterations (as determined by the isNCGstuck flag), this function will increment a counter nConsecutiveNcgStuck. If that counter exceeds a threshold of nConsecutiveNcgStuckThreshold, the usingNCG flag will be set to false, indicating that NCG should not be used. Otherwise, usingNCG remains true.

If the NCG optimizer is not currently stuck, the nConsecutiveNcgStuck counter is reset to 0 and usingNCG is set to true.

Returns: True if NCG should be used, false otherwise.

◆ reset_NCG_Rpi()

void OptimizationAlgorithm::reset_NCG_Rpi ( )

Reset NCG and Rpi to default.

Sets usingNCG to true, isNCGstuck to false, and usingRpi to true.

Member Data Documentation

◆ c1

double OptimizationAlgorithm::c1 = 1e-6

The c1 parameter in the NCG Wolfe condition.

◆ c2

double OptimizationAlgorithm::c2 = 0.001

The c2 parameter in the NCG Wolfe condition.

◆ energyDiffThreshold

double OptimizationAlgorithm::energyDiffThreshold = 1e-5

The energy difference threshold for convergence checking.

◆ forceDiffThreshold

double OptimizationAlgorithm::forceDiffThreshold = 1e-4

The mean force difference threshold for convergence checking.

◆ isCriteriaSatisfied

bool OptimizationAlgorithm::isCriteriaSatisfied = false

Flag indicating whether the optimization criteria have been met.

◆ isNCGstuck

bool OptimizationAlgorithm::isNCGstuck = false

Flag indicating whether the NCG optimizer has gotten stuck and cannot further minimize.

◆ nConsecutiveNcgStuck

int OptimizationAlgorithm::nConsecutiveNcgStuck = 0

The number of consecutive times the NCG optimizer has gotten stuck.

◆ nConsecutiveNcgStuckThreshold

int OptimizationAlgorithm::nConsecutiveNcgStuckThreshold = 3

The threshold of consecutive NCG stuck iterations before disabling it.

◆ stepThreshold

double OptimizationAlgorithm::stepThreshold = 1e-15

The lower bound for the step size in NCG.

◆ trialIterationInterval

int OptimizationAlgorithm::trialIterationInterval = 300

The number of iterations between trial step size adjustments.

◆ trialStepSize

double OptimizationAlgorithm::trialStepSize = 0.0

The step size used in the optimization algorithm.

◆ usingNCG

bool OptimizationAlgorithm::usingNCG = false

Flag indicating whether to use the nonlinear conjugate gradient optimizer.

◆ usingRpi

bool OptimizationAlgorithm::usingRpi = false

Flag indicating whether to use the R-pi adaptive method for regularization energy.

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

include/model/Model.hpp

Public Member Functions

Public Attributes