IOptimizer Class Reference

Base Optimizer Class. More...

#include <IOptimizer.h>

Inheritance diagram for IOptimizer:

List of all members.

Public Member Functions
	IOptimizer (EvalFuncPtr fPtr=NULL, unsigned int dn=PROBLEM_DIMENSIONS, unsigned int sn=NUM_SETS, unsigned int pn=NUM_SOLUTIONS)
	Create the optimizers. See setParams.
virtual bool	init ()
	Initialize the optimizer.
virtual void	free ()
	Free memory.
virtual void	optimize ()
	Launch the optimization.
void	setParams (EvalFuncPtr fPtr=NULL, unsigned int dn=PROBLEM_DIMENSIONS, unsigned int sn=NUM_SETS, unsigned int pn=NUM_SOLUTIONS)
	Set parameters used in the optimization.
void	setTerminationFlags (TERMINATION_FLAGS f)
	Set the criteria for stopping the optimization.
void	setGenerations (unsigned int gn)
	set maximum number of generations if TERMINATE_GENS is set (see setTerminationFlags)
void	setFunctionEvals (unsigned int fn)
	set maximum number of function evaluations if TERMINATE_EVALS is set (see setTerminationFlags)
void	setStoppingTime (float st)
	set limit of time if TERMINATE_TIME is set (see setTerminationFlags)
void	setStoppingFitness (float f)
	set limit fitness value if TERMINATE_FIT is set (see setTerminationFlags)
void	setStoppingFitness (unsigned int sId, float f)
	set limit fitness value to a single set if TERMINATE_FIT is set (see setTerminationFlags)
void	setNumberOfSets (unsigned int sn)
	set number of sets involved in the optimization process
void	setNumberOfSolutions (unsigned int pn)
	set number of solutions for each set
void	setProblemDimension (unsigned int dn)
	set the number of the dimensions of the problem
void	setEvalFuncPtr (EvalFuncPtr evalPtr)
	set the fitness function
void	setBounds (float2 bounds)
	set bounds for all sets
void	setBounds (unsigned int sId, float2 bounds)
	set bounds for a single set
void	setBounds (unsigned int sId, unsigned int dId, float2 bounds)
	set bounds for a single dimension in a single set
void	setVerbosity (LOG_LEVEL verb)
	set verbosity of the log
void	setOptimizationType (OPTIMUM ot)
	receives a OPTIMUM representing if the optimum of the function is the maximum or the minimum
void	setHostFitnessEvaluation (bool on)
	set this to true if the fitness evaluation will happen on host side: in this way the host pointer will always be updated
void	setFileOutput (std::string f)
	set the file chosen for the log
void	resetFileOutput ()
	set the stderr for log output
const SolutionSet *	getSolutionSet () const
	Get a constant copy of the solution set.
unsigned int	getNumberOfSets () const
	get number of sets
unsigned int	getNumberOfSolutions () const
	get number of solutions in a set
unsigned int	getProblemDimension () const
	get problem dimension
unsigned int	getFunctionEvals () const
	get number of evaluations used as a termination criteria
unsigned int	getCurrentEvals () const
	get number of evaluations performed in the optimization so far
unsigned int	getCurrentGeneration () const
	get number of generations performed in the optimization so far
TERMINATION_FLAGS	getTerminationFlags () const
	get termination flags
float	getElapsedTime () const
	get time passed from the beginning of optimization
virtual float *	getBestFitnesses ()
	get a pointer to best fitnesses of all sets
virtual float	getBestFitness (unsigned int sid)
	get value of the best fitness for set indicated by the index sid
virtual float *	getBestSolution ()
	get best solutions of all sets
virtual float *	getBestSolution (unsigned int sid)
	get best solution of the set indicated by the index sid
bool	isHostFitnessEvaluation () const
	returns true if the fitness evaluation will happen on host side
Protected Member Functions
virtual void	initSolutions (dim3 initializationGrid, dim3 initializationBlock)=0
	Initialize solutions.
virtual void	step (dim3 initializationGrid, dim3 initializationBlock)=0
	Generate solutions for subsequent fitnessEvaluation.
virtual void	fitnessEvaluation (dim3 initializationGrid, dim3 initializationBlock, bool first=false)=0
	Evaluate solutions.
virtual void	update (dim3 initializationGrid, dim3 initializationBlock, bool first=false)=0
	Update population according to the algorithm and the results of fitnessEvaluation.
virtual void	findBest ()
	Updates global best fitnesses, indexes in SolutionSet device memory.

---

Detailed Description

Base Optimizer Class.

To override this class you need to implement at least 4 methods:
initSolutions
step
fitnessEvaluation
update

Definition at line 71 of file IOptimizer.h.

---

Member Function Documentation

virtual void IOptimizer::fitnessEvaluation	(	dim3	initializationGrid,
		dim3	initializationBlock,
		bool	first = false
	)		[protected, pure virtual]

Evaluate solutions.

Parameters:

initializationGrid	organization of the CUDA blocks
initializationBlock	organization of the CUDA threads
first	it signals if it is the first evaluation of the optimization

Implemented in PSO_Optimizer, DE_Optimizer, and SW_Optimizer.

const SolutionSet* IOptimizer::getSolutionSet

(

)

const [inline]

Get a constant copy of the solution set.

Returns:

copy of the solution set

Definition at line 220 of file IOptimizer.h.

virtual void IOptimizer::initSolutions	(	dim3	initializationGrid,
		dim3	initializationBlock
	)		[protected, pure virtual]

Initialize solutions.

Parameters:

initializationGrid	organization of the CUDA blocks
initializationBlock	organization of the CUDA threads

Implemented in PSO_Optimizer, DE_Optimizer, and SW_Optimizer.

void IOptimizer::setParams	(	EvalFuncPtr	fPtr = NULL,
		unsigned int	dn = PROBLEM_DIMENSIONS,
		unsigned int	sn = NUM_SETS,
		unsigned int	pn = NUM_SOLUTIONS
	)		[inline]

Set parameters used in the optimization.

Parameters:

fPtr	as instance of EvalFuncPtr representing the fitness function
dn	problem dimension
sn	number of the sets that will run in the optimization
pn	number of solutions for each set

Definition at line 165 of file IOptimizer.h.

void IOptimizer::setTerminationFlags

(

TERMINATION_FLAGS

f

)

[inline]

Set the criteria for stopping the optimization.

More than one criterion can be set using | (or) operator. Possible values are TERMINATE_EVALS (use number of function evaluations), TERMINATE_GENS (number of generations), TERMINATE_FIT (value of fitness), TERMINATE_TIME (time)

Parameters:

f	the termination flags

Definition at line 182 of file IOptimizer.h.

virtual void IOptimizer::step	(	dim3	initializationGrid,
		dim3	initializationBlock
	)		[protected, pure virtual]

Generate solutions for subsequent fitnessEvaluation.

Parameters:

initializationGrid	organization of the CUDA blocks
initializationBlock	organization of the CUDA threads

Implemented in PSO_Optimizer, DE_Optimizer, and SW_Optimizer.

virtual void IOptimizer::update	(	dim3	initializationGrid,
		dim3	initializationBlock,
		bool	first = false
	)		[protected, pure virtual]

Update population according to the algorithm and the results of fitnessEvaluation.

Parameters:

initializationGrid	organization of the CUDA blocks
initializationBlock	organization of the CUDA threads
first	it signals if it is the first evaluation of the optimization

Implemented in PSO_Optimizer, DE_Optimizer, and SW_Optimizer.

---

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

• IOptimizer.h