npl::GradientOpt Class Reference

#include <gradient.h>

Inherits npl::Optimizer.

Public Member Functions
	GradientOpt (size_t dim, const ValFunc &valfunc, const GradFunc &gradfunc, const ValGradFunc &valgradfunc, const CallBackFunc &callback=noopCallback)
	Constructor for optimizer function. More...
	GradientOpt (size_t dim, const ValFunc &valfunc, const GradFunc &gradfunc, const CallBackFunc &callback=noopCallback)
	Constructor for optimizer function. More...
StopReason	optimize ()
	Perform optimization. More...

Static Public Member Functions
static std::string	explainStop (StopReason r)
	Provides a string that describes the stop reason. More...

Public Attributes
double	opt_maxstep
	Maximum step size, step will be rescaled to this length if it exceeds it after other scaling is complete. More...
double	opt_init_scale
	Initial scale to use during optimization, actual scale may differ due to other options. More...
double	opt_rdec_scale
	Multiply scale by this value after each iteration ( 0 < v < 1 ). Values <= 0 will be considered unused. More...
VectorXd	state_x
	State variable, set to initialize. More...
double	stop_G
	Stop when graient magnitde falls below this value. More...
double	stop_X
	Stop when step size drops below this value. More...
double	stop_F
	Stop when change in function value drops below this value. More...
double	stop_F_over
	Stop immediately when value goes above this. More...
double	stop_F_under
	Stop immediately when value goes below this. More...
int	stop_Its
	Stop after this many iterations (does not include linesearch) More...

Protected Attributes
ValGradFunc	m_compFG
GradFunc	m_compG
ValFunc	m_compF
CallBackFunc	m_callback

Detailed Description

Definition at line 36 of file gradient.h.

Constructor & Destructor Documentation

npl::GradientOpt::GradientOpt	(	size_t	dim,
		const ValFunc &	valfunc,
		const GradFunc &	gradfunc,
		const ValGradFunc &	valgradfunc,
		const CallBackFunc &	callback = noopCallback
	)

Constructor for optimizer function.

Parameters

dim	Dimensionality of state vector
valfunc	Function which computes the energy of the underlying mathematical function
gradfunc	Function which computes the gradient of energy in the underlying mathematical function
valgradfunc	Function which computes the both the energy and gradient in the underlying mathematical function
callback	Function which should be called at the end of each iteration (for instance, to debug)

npl::GradientOpt::GradientOpt	(	size_t	dim,
		const ValFunc &	valfunc,
		const GradFunc &	gradfunc,
		const CallBackFunc &	callback = noopCallback
	)

Constructor for optimizer function.

Parameters

dim	Dimensionality of state vector
valfunc	Function which computes the energy of the underlying mathematical function
gradfunc	Function which computes the gradient of energy in the underlying mathematical function
callback	Function which should be called at the end of each iteration (for instance, to debug)

Member Function Documentation

static std::string npl::Optimizer::explainStop ( StopReason  r )

staticinherited

Provides a string that describes the stop reason.

Parameters

r

StopReason npl::GradientOpt::optimize ( )

virtual

Perform optimization.

Returns

StopReason

Reimplemented from npl::Optimizer.

Member Data Documentation

CallBackFunc npl::Optimizer::m_callback

protectedinherited

Definition at line 245 of file opt.h.

ValFunc npl::Optimizer::m_compF

protectedinherited

Definition at line 244 of file opt.h.

ValGradFunc npl::Optimizer::m_compFG

protectedinherited

Definition at line 242 of file opt.h.

GradFunc npl::Optimizer::m_compG

protectedinherited

Definition at line 243 of file opt.h.

double npl::GradientOpt::opt_init_scale

Initial scale to use during optimization, actual scale may differ due to other options.

Definition at line 49 of file gradient.h.

double npl::GradientOpt::opt_maxstep

Maximum step size, step will be rescaled to this length if it exceeds it after other scaling is complete.

Definition at line 43 of file gradient.h.

double npl::GradientOpt::opt_rdec_scale

Multiply scale by this value after each iteration ( 0 < v < 1 ). Values <= 0 will be considered unused.

Definition at line 55 of file gradient.h.

VectorXd npl::Optimizer::state_x

inherited

State variable, set to initialize.

Definition at line 162 of file opt.h.

double npl::Optimizer::stop_F

inherited

Stop when change in function value drops below this value.

Definition at line 177 of file opt.h.

double npl::Optimizer::stop_F_over

inherited

Stop immediately when value goes above this.

Definition at line 182 of file opt.h.

double npl::Optimizer::stop_F_under

inherited

Stop immediately when value goes below this.

Definition at line 187 of file opt.h.

double npl::Optimizer::stop_G

inherited

Stop when graient magnitde falls below this value.

Definition at line 167 of file opt.h.

int npl::Optimizer::stop_Its

inherited

Stop after this many iterations (does not include linesearch)

Definition at line 192 of file opt.h.

double npl::Optimizer::stop_X

inherited

Stop when step size drops below this value.

Definition at line 172 of file opt.h.

---

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

• /home/micahc/coding/npl/deps/optimizers/lib/gradient.h