A class that serves as the main interface for evolutionary algorithms in Grail. More...

#include <EAOptimizer.hh>

Public Member Functions
	EAOptimizer (std::vector< std::unique_ptr< Individual >> &initialPopulation, size_t maxEpochCount, std::unique_ptr< Crossover > crossover=std::make_unique< Crossover >(), std::unique_ptr< Mutation > mutation=std::make_unique< Mutation >(), std::unique_ptr< Selection > selection=std::make_unique< Selection >(), std::mt19937_64::result_type seed=std::random_device{}())
	The EA algorithm will maintain the size of the initial population after the selection phase. More...

void	RunInteractive (std::function< void(Individual &individualToEvaluate)> onEvaluateNeeded, std::function< void(void)> onFinished=nullptr, std::function< void(void)> onNewEpoch=nullptr)
	Advances the computations of the evolutionary algorithm (EA). This is the main method to run it. This method will call provided functions on specific events. More...

size_t	GetPopulationSize () const
	PopulationSize - returns the number of individuals in the current population. More...

std::vector< std::unique_ptr< Individual > >	GetPopulation () const

void	SetElitism (double rate, ElitismType type)
	SetElitism - sets how many individuals advance to the next generation in an unmodified form and how they are chosen. These are called elite individuals. More...

double	GetElitismRate () const

ElitismType	GetElitismType () const

float	CalculatePopulationDiversity () const
	Returns the diversity ratio as a number [0,1]. The lower the value, the population is more homogeneous. More...

size_t	CalculatePopulationIdenticalFrontCount () const
	Finds the most frequent individuals (which are identical to each other) and returns the number of them. More...

float	CalculatePopulationIdenticalFrontRatio () const
	Returns the number of the most frequent identical individuals divided by the population size. More...

const Individual &	GetBestIndividual () const
	GetBestIndividual - gets the individual with the highest fitness value from the population. More...

size_t	GetEpochNumber () const

size_t	GetMaxEpochCount () const

Crossover &	GetCrossover () const
	GetCrossover - gets the crossover instance that is used by this evolutionary algorithm. More...

Mutation &	GetMutation () const
	GetMutation - gets the mutation instance that is used by this evolutionary algorithm. More...

Selection &	GetSelection () const
	GetSelection - gets the selection instance that is used by this evolutionary algorithm. More...

void	SerializeState (const std::string &filename) const
	Serializes the inner state of the evolutionary algorithm, including the current population. You can load the serialized state again by calling the 'DeserializeState' method. More...

bool	DeserializeState (const std::string &filename)
	Loads a previously serialized state of the evolutionary algorithm. You can, for instance, get the results or continue calculations. More...

Friends
std::ostream &	operator<< (std::ostream &out, const EAOptimizer &eaOptimizer)
	Appends to stream the serialized state of the evolutionary algorithm including the whole population and basic statistics about the population. More...

Detailed Description

A class that serves as the main interface for evolutionary algorithms in Grail.

Constructor & Destructor Documentation

◆ EAOptimizer()

grail::evolution::EAOptimizer::EAOptimizer	(	std::vector< std::unique_ptr< Individual >> &	initialPopulation,
		size_t	maxEpochCount,
		std::unique_ptr< Crossover >	crossover = `std::make_unique<Crossover>()`,
		std::unique_ptr< Mutation >	mutation = `std::make_unique<Mutation>()`,
		std::unique_ptr< Selection >	selection = `std::make_unique<Selection>()`,
		std::mt19937_64::result_type	seed = `std::random_device{}()`
	)

The EA algorithm will maintain the size of the initial population after the selection phase.

Parameters

initialPopulation	- the initial population to the EA algorithm. For example, you can create one individual and call GetRandomRealizations() on it.
maxEpochCount	- the maximum number of full iterations after which the algorithm will stop making any further optimizations.
crossover	- a crossover struct instance that encapsulates the crossover (recombination) procedure of evolutionary algorithm.
mutation	- a mutation struct instance that encapsulates the mutation procedure of evolutionary algorithm.
selection	a selection class instance that encapsulates the selection procedure of evolutionary algorithm.
seed	- the seed to the random numbers generator.

Member Function Documentation

◆ CalculatePopulationDiversity()

float grail::evolution::EAOptimizer::CalculatePopulationDiversity ( ) const

Returns the diversity ratio as a number [0,1]. The lower the value, the population is more homogeneous.

◆ CalculatePopulationIdenticalFrontCount()

size_t grail::evolution::EAOptimizer::CalculatePopulationIdenticalFrontCount ( ) const

Finds the most frequent individuals (which are identical to each other) and returns the number of them.

◆ CalculatePopulationIdenticalFrontRatio()

float grail::evolution::EAOptimizer::CalculatePopulationIdenticalFrontRatio ( ) const

Returns the number of the most frequent identical individuals divided by the population size.

◆ DeserializeState()

bool grail::evolution::EAOptimizer::DeserializeState ( const std::string & filename )

Loads a previously serialized state of the evolutionary algorithm.
You can, for instance, get the results or continue calculations.

Parameters

filename - the path where the state has been saved.

Returns: - a boolean value that tells if the deserialization was successful.

◆ GetBestIndividual()

const Individual & grail::evolution::EAOptimizer::GetBestIndividual ( ) const

GetBestIndividual - gets the individual with the highest fitness value from the population.

◆ GetCrossover()

Crossover & grail::evolution::EAOptimizer::GetCrossover ( ) const

GetCrossover - gets the crossover instance that is used by this evolutionary algorithm.

◆ GetElitismRate()

double grail::evolution::EAOptimizer::GetElitismRate ( ) const

Returns: the ElitismRate - the portion of elite individuals to the whole population.

◆ GetElitismType()

ElitismType grail::evolution::EAOptimizer::GetElitismType ( ) const

Returns: the ElitismType that determines how elite individuals are chosen. See the description of ElitismType.

◆ GetEpochNumber()

size_t grail::evolution::EAOptimizer::GetEpochNumber ( ) const

Returns: the current epoch (iteration) number.

◆ GetMaxEpochCount()

size_t grail::evolution::EAOptimizer::GetMaxEpochCount ( ) const

Returns: the number of epochs (iterations) of EA, after which the algorithm will stop.

◆ GetMutation()

Mutation & grail::evolution::EAOptimizer::GetMutation ( ) const

GetMutation - gets the mutation instance that is used by this evolutionary algorithm.

◆ GetPopulation()

std::vector< std::unique_ptr< Individual > > grail::evolution::EAOptimizer::GetPopulation ( ) const

Returns: A vector, sorted by fitness, that contains deep clones of the current population evolved by the algorithm.

◆ GetPopulationSize()

size_t grail::evolution::EAOptimizer::GetPopulationSize ( ) const

PopulationSize - returns the number of individuals in the current population.

Returns: the size of the current population.

◆ GetSelection()

Selection & grail::evolution::EAOptimizer::GetSelection ( ) const

GetSelection - gets the selection instance that is used by this evolutionary algorithm.

◆ RunInteractive()

void grail::evolution::EAOptimizer::RunInteractive	(	std::function< void(Individual &individualToEvaluate)>	onEvaluateNeeded,
		std::function< void(void)>	onFinished = `nullptr`,
		std::function< void(void)>	onNewEpoch = `nullptr`
	)

Advances the computations of the evolutionary algorithm (EA). This is the main method to run it. This method will call provided functions on specific events.

Parameters

onEvaluateNeeded	- when EA needs an individual to have it fitness evaluated, it will invoke this callback function with this individual as a parameter.
onFinished	- this callback will be invoked when the algorithm has reached its maximum number of iterations.
onNewEpoch	- this callback will be invoked when a new epoch (iteration) starts. An epoch is one round of evaluating a population and applying genetic operators on it.

◆ SerializeState()

void grail::evolution::EAOptimizer::SerializeState ( const std::string & filename ) const

Serializes the inner state of the evolutionary algorithm, including the current population.
You can load the serialized state again by calling the 'DeserializeState' method.

Parameters

filename - the path where the state will be saved. If the file exists, it will be overridden.

◆ SetElitism()

void grail::evolution::EAOptimizer::SetElitism	(	double	rate,
		ElitismType	type
	)

SetElitism - sets how many individuals advance to the next generation in an unmodified form and how they are chosen. These are called elite individuals.

Parameters

rate	- the portion of elite individuals to the whole population.
type	- determines how elite individuals are chosen. See the description of ElitismType.

Friends And Related Function Documentation

◆ operator<<

std::ostream& operator<<	(	std::ostream &	out,
		const EAOptimizer &	eaOptimizer
	)

friend

Appends to stream the serialized state of the evolutionary algorithm including the whole population and basic statistics about the population.

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

GrailEvolution/EAOptimizer.hh
GrailEvolution/EAOptimizer.cpp

Public Member Functions

Friends

Detailed Description

Constructor & Destructor Documentation

◆ EAOptimizer()

Member Function Documentation

◆ CalculatePopulationDiversity()

◆ CalculatePopulationIdenticalFrontCount()

◆ CalculatePopulationIdenticalFrontRatio()

◆ DeserializeState()

◆ GetBestIndividual()

◆ GetCrossover()

◆ GetElitismRate()

◆ GetElitismType()

◆ GetEpochNumber()

◆ GetMaxEpochCount()

◆ GetMutation()

◆ GetPopulation()

◆ GetPopulationSize()

◆ GetSelection()

◆ RunInteractive()

◆ SerializeState()

◆ SetElitism()

Friends And Related Function Documentation

◆ operator<<