#include <garp.hh>

Inheritance diagram for Garp:

Collaboration diagram for Garp:

Public Member Functions
	Garp ()
virtual	~Garp ()
int	initialize ()
int	iterate ()
int	done () const
float	getProgress () const
Scalar	getValue (const Sample &x) const
int	getConvergence (Scalar *const val) const
int	getGeneration ()
void	deleteTempDataMembers ()
Protected Member Functions
virtual void	_getConfiguration (ConfigurationPtr &) const
virtual void	_setConfiguration (const ConstConfigurationPtr &)
Private Member Functions
void	cacheSamples (const SamplerPtr &, OccurrencesPtr &, int resamples)
void	colonize (GarpRuleSet *ruleset, int numRules)
void	evaluate (GarpRuleSet *ruleset)
void	keepFittest (GarpRuleSet source, GarpRuleSet target, PerfIndex perfIndex)
void	select (GarpRuleSet source, GarpRuleSet target, double gapsize)
void	mutate (GarpRuleSet *ruleset)
void	crossover (GarpRuleSet *ruleset)
Private Attributes
int	_max_gen
int	_popsize
int	_resamples
double	_acc_limit
double	_conv_limit
double	_mortality
double	_significance
double	_crossover_rate
double	_mutation_rate
double	_gapsize
GarpRuleSet *	_fittest
GarpRuleSet *	_offspring
BioclimHistogram	_bioclimHistogram
Regression	_regression
OccurrencesPtr	_cachedOccs
double	_convergence
int	_improvements
int	_curr_heur_count [5]
int	_prev_heur_count [5]
int	_gen
float	_maxProgress

Detailed Description

Implementation of GARP: Genetic Algorithm for Rule-set Production

Definition at line 57 of file garp.hh.

Constructor & Destructor Documentation

Garp::Garp ( )

Definition at line 231 of file garp.cpp.

References _acc_limit, _conv_limit, _convergence, _crossover_rate, _curr_heur_count, _fittest, _gapsize, _gen, _improvements, _max_gen, _maxProgress, _mortality, _mutation_rate, AlgorithmImpl::_normalizerPtr, _offspring, _popsize, _prev_heur_count, _resamples, and _significance.

Garp::~Garp ( ) [virtual]

Definition at line 270 of file garp.cpp.

References _fittest, and _offspring.

Member Function Documentation

void Garp::_getConfiguration ( ConfigurationPtr & config ) const [protected, virtual]

Reimplemented from AlgorithmImpl.

Definition at line 444 of file garp.cpp.

References _acc_limit, _crossover_rate, _fittest, _gapsize, _gen, _mortality, _mutation_rate, _significance, done(), GarpRuleSet::get(), GarpRule::getChrom1(), GarpRule::getChrom2(), GarpRule::getPerformanceArray(), GarpRule::getPrediction(), GarpRuleSet::numRules(), and GarpRule::type().

Here is the call graph for this function:

void Garp::_setConfiguration ( const ConstConfigurationPtr & config ) [protected, virtual]

Reimplemented from AlgorithmImpl.

Definition at line 489 of file garp.cpp.

References _acc_limit, _crossover_rate, _fittest, _gapsize, _gen, _mortality, _mutation_rate, _offspring, _popsize, _significance, GarpRuleSet::add(), Log::error(), AlgorithmImpl::getParameter(), Log::instance(), and Sample::size().

Here is the call graph for this function:

void Garp::cacheSamples	(	const SamplerPtr &	sampler,
		OccurrencesPtr &	cachedOccs,
		int	resamples
	)		`[private]`

Definition at line 331 of file garp.cpp.

References GeoTransform::getDefaultCS(), and OccurrencesImpl::reserve().

Referenced by initialize().

Here is the call graph for this function:

void Garp::colonize	(	GarpRuleSet *	ruleset,
		int	numRules
	)		`[private]`

Fill a rule set with newly generated rules

Parameters:

ruleset,:	rule set where rules will be added to
numRules	number of rules to be added to the rule set.

Definition at line 656 of file garp.cpp.

References _bioclimHistogram, _regression, AlgorithmImpl::_samp, GarpRuleSet::add(), Random::get(), GarpRuleSet::numRules(), and GarpRule::setPrediction().

Referenced by initialize(), and iterate().

Here is the call graph for this function:

void Garp::crossover ( GarpRuleSet * ruleset ) [private]

Perform crossing over with rules in a rule set

Parameters:

ruleset Rule set containing rules to be crossed over

Definition at line 793 of file garp.cpp.

References _crossover_rate, AlgorithmImpl::_samp, GarpRule::crossover(), Random::get(), GarpRuleSet::get(), and GarpRuleSet::numRules().

Referenced by iterate().

Here is the call graph for this function:

void Garp::deleteTempDataMembers ( )

Deletes data structures that are not needed once the model is generated. Implemented to save some memory when clients have to store too many Garp objects.

Definition at line 818 of file garp.cpp.

References _offspring.

int Garp::done ( ) const [virtual]

Check if the model generation is completed (e.g. convergence point has been met.

Returns:: Implementation specific but usually 1 for completion.

Reimplemented from AlgorithmImpl.

Definition at line 419 of file garp.cpp.

References _conv_limit, _convergence, _gen, and _max_gen.

Referenced by _getConfiguration(), getProgress(), and iterate().

void Garp::evaluate ( GarpRuleSet * ruleset ) [private]

Evaluate rules from a rule set based on their performance to predict points provided by a sampler object.

Parameters:

ruleset,: rule set to be tested.

Definition at line 640 of file garp.cpp.

References _cachedOccs, GarpRule::evaluate(), GarpRuleSet::get(), and GarpRuleSet::numRules().

Referenced by iterate().

Here is the call graph for this function:

int Garp::getConvergence ( Scalar *const val ) const [virtual]

Returns a value that represents the convergence of the algorithm expressed as a number between 0 and 1 where 0 represents model completion.

Returns:

Parameters:

val

Reimplemented from AlgorithmImpl.

Definition at line 435 of file garp.cpp.

References _convergence.

int Garp::getGeneration ( ) [inline]

Definition at line 111 of file garp.hh.

References _gen.

float Garp::getProgress ( ) const [virtual]

Return progress so far

Reimplemented from AlgorithmImpl.

Definition at line 401 of file garp.cpp.

References _conv_limit, _convergence, _gen, _max_gen, _maxProgress, and done().

Here is the call graph for this function:

Scalar Garp::getValue ( const Sample & x ) const [virtual]

This method is used when projecting the model.

Note:: This method is inherited from the Algorithm class

Returns:

Parameters:

x	a pointer to a vector of openModeller Scalar type (currently double). The vector should contain values looked up on the environmental variable layers into which the mode is being projected.

Implements AlgorithmImpl.

Definition at line 427 of file garp.cpp.

References _fittest, and GarpRuleSet::getValue().

Here is the call graph for this function:

int Garp::initialize ( ) [virtual]

Initialize model by colonizing it with new rules.

Note:: This method is inherited from the Algorithm class

Returns:: 0 on error

Implements AlgorithmImpl.

Definition at line 290 of file garp.cpp.

References _bioclimHistogram, _cachedOccs, _conv_limit, _fittest, _max_gen, _offspring, _popsize, _regression, _resamples, AlgorithmImpl::_samp, cacheSamples(), Regression::calculateParameters(), colonize(), Log::error(), AlgorithmImpl::getParameter(), BioclimHistogram::initialize(), and Log::instance().

Here is the call graph for this function:

int Garp::iterate ( ) [virtual]

Start model execution (build the model). This method is the main body of the Genetic Algorithm. The steps during the algorithm are: 1) Evaluate current population testing the rules against data points 2) Select fittest individuals from population based on significance 3) If reached any stop condition (# of iterations or convergence limit) then output best result and stop 4) Else perform reproduction, generating a new offspring from current population 5) Repeat from step 1.

Returns:: 0 on error

Reimplemented from AlgorithmImpl.

Definition at line 351 of file garp.cpp.

References _fittest, _gapsize, _gen, _offspring, _popsize, _significance, colonize(), crossover(), defaultPerfIndex, done(), evaluate(), GarpRuleSet::filter(), keepFittest(), mutate(), GarpRuleSet::performanceSummary(), select(), and GarpRuleSet::trim().

Here is the call graph for this function:

void Garp::keepFittest	(	GarpRuleSet *	source,
		GarpRuleSet *	target,
		PerfIndex	perfIndex
	)		`[private]`

Keep fittest individuals from source rule set storing them in target rule set in descending order of fitness.

Parameters:

source	Pointer to the source rule set where rules will be selected from (usually the previous generated population)
target	Pointer to the target rule set where fittest rules will be transferred to (usually the fittest rule set)
perfIndex	Index of value on performance array that will be used to select individuals.

Note:: : This method produces a side effect that is to update the statistics about heuristic operators performance (that will be used to adjust probabilities of application of those operators on future generations) and overall rule performance that will be used to compute the changes to the convergence value.

Definition at line 579 of file garp.cpp.

References _convergence, _improvements, GarpRule::clone(), GarpRuleSet::findSimilar(), GarpRuleSet::get(), GarpRule::getPerformance(), GarpRuleSet::insert(), GarpRuleSet::numRules(), and GarpRuleSet::remove().

Referenced by iterate().

Here is the call graph for this function:

void Garp::mutate ( GarpRuleSet * ruleset ) [private]

Mutate rules in a rule set

Parameters:

ruleset Rule set containing rules to be mutated

Definition at line 780 of file garp.cpp.

References _gen, GarpRuleSet::get(), GarpRule::mutate(), and GarpRuleSet::numRules().

Referenced by iterate().

Here is the call graph for this function:

void Garp::select	(	GarpRuleSet *	source,
		GarpRuleSet *	target,
		double	gapsize
	)		`[private]`

Select fittest individuals and store them into target ruleset

Parameters:

source	Pointer to the rule set containing parents (individuals that will provide genes to create new individuals).
target	Pointer to the target rule set where offspring will be transferred to.
gapsize	Proportion of contribution from the objBest archive set to the new population.

Definition at line 699 of file garp.cpp.

References _popsize, GarpRuleSet::add(), GarpRuleSet::clear(), GarpRule::clone(), defaultPerfIndex, error, Log::error(), GarpRule::forceEvaluation(), Random::get(), GarpRuleSet::get(), GarpRule::getPerformance(), Log::instance(), GarpRuleSet::numRules(), and GarpRuleSet::performanceSummary().