Encodings

Encodings#

namespace gapp

struct GaTraits#

#include <core/ga_traits.hpp>

template<typename GeneType> struct GaTraits#

Traits class describing some attributes of each GA class.

When defining a new genetic algorithm class that inherits from the GA class, this class must be specialized for the gene type T used before the derived class is declared, and must have the following members:

DefaultCrossover type (must be default constructible)
DefaultMutation type (must be constructible using the return value of defaultMutationRate)
defaultMutationRate(size_t chrom_len) -> Probability (static member function)

The following gene types are reserved for the GAs already implemented in the library and can’t be used as the gene type of new encodings:

std::uint8_t
std::size_t
int
double

Example:

template<>
struct GaTraits<MyGeneType>
{
  using DefaultCrossover = MyCrossoverType;
  using DefaultMutation  = MyMutationType;
  static Probability defaultMutationRate(size_t chromosome_size) { return 0.01; }
};

See also

bool is_bounded#

#include <core/ga_traits.hpp>

template<typename GeneType> constexpr bool gapp::is_bounded = false#

Type trait to check if a particular gene type is bounded or not. This variable should be specialized for custom gene types if they are bounded, otherwise they will be treated as if they were not bounded.

Example:

template<> inline constexpr bool is_bounded<MyGene> = true;

class GA#

#include <core/ga_base.hpp>

template<typename T> class GA : public gapp::GaInfo #

The base class used for all of the genetic algorithms. Contains all of the encoding/gene specific information of the GA in addition to the non-encoding dependent parts contained by GaInfo.

This class should be used as the base class of new genetic algorithms. It has 2 virtual functions that should be implemented in the derived classes:

initialize (optional) : Perform some additional initialization specific to the derived class.
generateCandidate : Returns a (randomly) generated candidate solution. Used to create the initial population.

See also

GaTraits, is_bounded

Note

Before declaring a class that is derived from this class, the GaTraits struct should be specialized for the particular gene type that is used. If the gene type is bounded, the is_bounded variable template should also be specialized.

Template Parameters:: T – The gene type used for the candidate’s chromosomes.

Public Types

using GeneType = T #: The gene type of the candidates.

using CrossoverCallable = std::function<CandidatePair<T>(const GA&, const Candidate<T>&, const Candidate<T>&)>#

The general callable type that can be used as a crossover method in the GA (when not using a crossover method that is derived from crossover::Crossover). The function takes two candidate solutions (parents), and returns two candidates generated from these (children).

See also

crossover_method

using MutationCallable = std::function<void(const GA&, const Candidate<T>&, Chromosome<T>&)>#

The general callable type that can be used as a mutation method in the GA (when not using a mutation method derived from mutation::Mutation). The function takes a candidate solution, and changes this solution’s chromosome.

See also

mutation_method

using RepairCallable = std::function<Chromosome<T>(const GA&, const Chromosome<T>&)>#

The general callable type that can be used as a repair function in the GA. The function takes a candidate solution, and performs some operation on it.

See also

repair_function

Public Functions

GA(Positive<size_t> population_size = DEFAULT_POPSIZE)#

Create a genetic algorithm using the default genetic operators.

The algorithm used will be deduced from the number of objectives of the fitness function (single- or multi-objective), along with the mutation probability used, which will be deduced from the chromosome length.

Parameters:: population_size – The number of candidates in the population. Must be at least 1.

GA( Positive<size_t> population_size, std::unique_ptr<algorithm::Algorithm> algorithm, )#

Create a genetic algorithm using the default genetic operators. The mutation probability used will be deduced from the chromosome length.

Parameters:

population_size – The number of candidates in the population. Must be at least 1.
algorithm – The algorithm to use. The default algorithm will be used if it’s a nullptr.

GA( Positive<size_t> population_size, std::unique_ptr<crossover::Crossover<T>> crossover, std::unique_ptr<mutation::Mutation<T>> mutation, std::unique_ptr<stopping::StopCondition> stop_condition = std::make_unique<stopping::NoEarlyStop>(), )#

Create a genetic algorithm using the specified operators. The algorithm used will be deduced from the number of objectives of the fitness function (single- or multi-objective).

Parameters:

population_size – The number of candidates in the population. Must be at least 1.
crossover – The crossover operator to use. Can’t be a nullptr.
mutation – The mutation operator to use. Can’t be a nullptr.
stop_condition – The early-stop condition to use. No early-stopping will be used if it’s a nullptr.

GA( Positive<size_t> population_size, std::unique_ptr<algorithm::Algorithm> algorithm, std::unique_ptr<crossover::Crossover<T>> crossover, std::unique_ptr<mutation::Mutation<T>> mutation, std::unique_ptr<stopping::StopCondition> stop_condition = std::make_unique<stopping::NoEarlyStop>(), )#

Create a genetic algorithm using the specified algorithm and operators.

Parameters:

population_size – The number of candidates in the population. Must be at least 1.
algorithm – The algorithm to use. The default algorithm will be used if it’s a nullptr.
crossover – The crossover operator to use. Can’t be a nullptr.
mutation – The mutation operator to use. Can’t be a nullptr.
stop_condition – The early-stop condition to use. No early-stopping will be used if it’s a nullptr.

template<typename AlgorithmType> requires std::derived_from<AlgorithmType, algorithm::Algorithm> GA( Positive<size_t> population_size, AlgorithmType algorithm, )#

Create a genetic algorithm using the default genetic operators. The mutation probability will be deduced from the chromosome length.

Parameters:

population_size – The number of candidates in the population. Must be at least 1.
algorithm – The algorithm to use.

template<typename CrossoverType, typename MutationType, typename StoppingType = stopping::NoEarlyStop> requires std::derived_from<CrossoverType, crossover::Crossover<T>> && std::derived_from<MutationType, mutation::Mutation<T>> && std::derived_from<StoppingType, stopping::StopCondition> GA( Positive<size_t> population_size, CrossoverType crossover, MutationType mutation, StoppingType stop_condition = {}, )#

Create a genetic algorithm using the specified operators. The algorithm used will be deduced from the number of objectives of the fitness function (single- or multi-objective).

Parameters:

population_size – The number of candidates in the population. Must be at least 1.
crossover – The crossover operator to use.
mutation – The mutation operator to use.
stop_condition – The early-stop condition to use.

template<typename AlgorithmType, typename CrossoverType, typename MutationType, typename StoppingType = stopping::NoEarlyStop> requires std::derived_from<AlgorithmType, algorithm::Algorithm> && std::derived_from<CrossoverType, crossover::Crossover<T>> && std::derived_from<MutationType, mutation::Mutation<T>> && std::derived_from<StoppingType, stopping::StopCondition> GA( Positive<size_t> population_size, AlgorithmType algorithm, CrossoverType crossover, MutationType mutation, StoppingType stop_condition = {}, )#

Create a genetic algorithm using the specified algorithm and operators.

Parameters:

population_size – The number of candidates in the population. Must be at least 1.
algorithm – The algorithm to use.
crossover – The crossover operator to use.
mutation – The mutation operator to use.
stop_condition – The early-stop condition to use.

inline virtual const FitnessFunctionBase<T> *fitness_function( ) const & noexcept final#

Returns:: The fitness function used. A nullptr is returned if no fitness function is set.

inline virtual size_t chrom_len() const noexcept final#

Returns:: The chromosome length used. Returns 0 if the chromosome length is unspecified (ie. no fitness function was set yet).

inline const BoundsVector<T> &gene_bounds( ) const noexcept# requires (is_bounded<T>)

See also

Bounds

Returns:: The lower and upper bounds for each of the chromosomes’ genes (the ranges are inclusive).

template<typename F> inline requires std::derived_from< F, crossover::Crossover< T > > void crossover_method (F f)

Set the crossover method the GA will use. The crossover method should be thread-safe if parallel execution is enabled (true by default).

Note

As the crossover probability is associated with a crossover method, setting a new crossover method will override any previous crossover probability set using crossover_rate().

Parameters:: f – The crossover method to use.

inline void crossover_method(std::unique_ptr<crossover::Crossover<T>> f)#

Set the crossover method the GA will use. The crossover method should be thread-safe if parallel execution is enabled (true by default).

Note

As the crossover probability is associated with a crossover method, setting a new crossover method will override any previous crossover probability set using crossover_rate().

Parameters:: f – The crossover method to use. Can’t be a nullptr.

inline void crossover_method(CrossoverCallable f)#

Set the crossover method the GA will use. The crossover method should be thread-safe if parallel execution is enabled (true by default).

See also

CrossoverCallable

Note

As the crossover probability is associated with a crossover method, setting a new crossover method will override any previous crossover probability set using crossover_rate(). The crossover probability that will be used for simple callables will be the default crossover rate by default, but it can be changed using crossover_rate().

Parameters:: f – The crossover method to use.

const crossover::Crossover<T> &crossover_method() const & noexcept#

Returns:: The crossover operator used by the GA.

inline virtual void crossover_rate(Probability pc) noexcept final#

Set the crossover rate of the crossover operator used.

See also

crossover_method

Note

This crossover rate will be set for the current crossover method, and it will be changed if a new crossover method is set.

Parameters:: pc – The crossover probability. Must be in the closed interval [0.0, 1.0].

inline virtual Probability crossover_rate() const noexcept final#

Returns:: The crossover rate set for the crossover operator.

template<typename F> inline requires std::derived_from< F, mutation::Mutation< T > > void mutation_method (F f)

Set the mutation method the GA will use. The mutation method should be thread-safe if parallel execution is enabled (true by default).

Note

As the mutation probability is associated with a mutation method, setting a new mutation method will override any previous mutation rate set using mutation_rate().

Parameters:: f – The mutation method to use.

inline void mutation_method(std::unique_ptr<mutation::Mutation<T>> f)#

Set the mutation method the GA will use. The mutation method should be thread-safe if parallel execution is enabled (true by default).

Note

As the mutation probability is associated with a mutation method, setting a new mutation method will override any previous mutation rate set using mutation_rate().

Parameters:: f – The mutation method to use. Can’t be a nullptr.

inline void mutation_method(MutationCallable f)#

Set the mutation method the GA will use. The mutation method should be thread-safe if parallel execution is enabled (true by default).

See also

MutationCallable

Note

As the mutation probability is associated with a mutation method, setting a new mutation method will override any previous mutation rate set using mutation_rate(). The mutation probability that will be used for simple callables will be the default (deduced) mutation rate by default, but it can be changed using mutation_rate().

Parameters:: f – The mutation method to use.

const mutation::Mutation<T> &mutation_method() const & noexcept#

Returns:: The mutation operator used by the GA.

inline virtual void mutation_rate(Probability pm) noexcept final#

Set the mutation rate of the mutation operator.

See also

mutation_method

Note

This mutation rate will be set for the current mutation method, and it will be changed if a new mutation method is set.

Parameters:: pm – The mutation probability. Must be in the closed interval [0.0, 1.0].

inline virtual Probability mutation_rate() const noexcept final#

Returns:: The mutation probability set for the mutation operator.

inline void repair_function(RepairCallable f)#

Set a repair function for the GA. This function will be invoked for each chromosome in every generation, after the chromosomes were mutated.

Using a repair function is optional, and no repair function is used in the GA by default. A nullptr can be used as the argument to this function to disable the use of a repair function.

The repair function should be thread-safe if parallel execution is enabled (true by default).

Parameters:: f – The repair function to use. Allowed to be a nullptr.

inline const Candidates<T> &solutions() const & noexcept#

Returns:: The pareto-optimal solutions found by the GA. These are the optimal solutions of the last generation’s population if keep_all_optimal_solutions() is not set, otherwise it contains every optimal solution found during the run (the solution set is updated in every generation in this case).

inline const Population<T> &population() const & noexcept#

Returns:: The current population of the algorithm. This is the entire population, which may contain non optimal solutions, unlike the population returned by solutions().

Candidates<T> solve( std::unique_ptr<FitnessFunctionBase<T>> fitness_function, size_t generations, Population<T> initial_population = {}, )# requires (!is_bounded<T>)

Find the maximum of a fitness function using the genetic algorithm.

Specifying an initial population is optional and it may have any number of solutions, but only the first population_size() elements will be used if it’s greater than the population size set for the GA. If it has less candidates than the population size, the rest of the candidates will be generated using the generateCandidate() method.

Parameters:

fitness_function – The fitness function to find the maximum of. Can’t be a nullptr.
generations – The maximum number of generations the run can last for.
initial_population – The starting population to use as the first generation of the GA.

Returns: