Ga

Ga

class gaggle.ga.ga.GA(population_manager: PopulationManager = None, ga_args: GAArgs = None, selection: Selection = None, crossover: Crossover = None, mutation: Mutation = None, problem_args: ProblemArgs = None, sys_args: SysArgs = None, outdir_args: OutdirArgs = None, individual_args: IndividualArgs = None, problem: Problem = None)

Bases: object

The parent class for any GA. Is used to store all the information related to the GA algorithm and organize the order of the operators.

abstract train(*args, **kwargs)

Function used to evolve the GA. Needs to be overwritten.

Ga_factory

class gaggle.ga.ga_factory.GAFactory

Bases: object

Factory that generates pre-existing available GA algorithms. GAFactory.gas stores said GAs as a dictionary with their name as key and the uninitialized GA object as value.

classmethod from_ga_args(population_manager: PopulationManager = None, ga_args: GAArgs = None, selection: Selection = None, crossover: Crossover = None, mutation: Mutation = None, problem_args: ProblemArgs = None, sys_args: SysArgs = None, outdir_args: OutdirArgs = None, individual_args: IndividualArgs = None, problem: Problem = None) → GA

Initializes the requested GA from the dictionary of available GAs.

This is done by using the attribute ga_args.ga_name as the lookup key to GAFactory.gas.

Parameters:

• population_manager –

• ga_args –

• selection –

• crossover –

• mutation –

• problem_args –

• sys_args –

• outdir_args –

• individual_args –

• problem –

Returns:

An initialized GA class object.

gas = {'simple': <class 'gaggle.ga.simple_ga.SimpleGA'>}

classmethod update(key, ga)

Add a new GA to the dictionary of GAs that can be created.

It is added to GAFactory.gas

Parameters:

• key – dataset name that will be used as the dictionary lookup key

• ga – GA class object, it needs to not be already initialized

Simple_ga

class gaggle.ga.simple_ga.SimpleGA(population_manager: PopulationManager = None, ga_args: GAArgs = None, selection: Selection = None, crossover: Crossover = None, mutation: Mutation = None, problem_args: ProblemArgs = None, sys_args: SysArgs = None, outdir_args: OutdirArgs = None, individual_args: IndividualArgs = None, problem: Problem = None)

Bases: GA

Implements a Simple Genetic Algorithm following Mitchell.

display_metrics(display_train: bool = True, display_test: bool = True)

Displays the metrics computed and stored in self.saved_metrics. Only displays and saves the metrics whose keys are in self.metrics_to_plot. The graphs are also saved to a file in the output folder specified in outdir_args.

Parameters:

• display_train –

• display_test –

Returns:

get_fitness_metric(fitness: dict[slice(<class 'int'>, <class 'float'>, None)], save: bool = False, mode: str = 'train')

Outputs basic fitness metrics for a population, like avg fitness, best & worst fitness :param fitness: dictionary of fitness :param save: whether to save the metrics :param mode: what to save :return: metrics: dictionary of metrics

save_individual(individual: Individual, fitness)

Saves the individual provided as well as its fitness to the folder path specified in the outdir_args as ‘best.pt’.

Parameters:

• individual –

• fitness –

Returns:

save_metrics(save_train: bool = True, save_test: bool = True)

Saves the metrics computed and stored in self.saved_metrics. The metrics are saved to files in the output folder specified in outdir_args.

Parameters:

• save_train –

• save_test –

Returns:

test(save: bool = True)

Tests population by computing the fitness and its associated metrics without updating the population manager.

Parameters:

save –

Returns:

train(test: bool = True, callbacks: List[Callable] = None, display_train_metrics: bool = True, display_test_metrics: bool = True)

Call to begin the training process of the population using the arguments stored in this SimpleGA object.

Parameters:

• test –

• callbacks –

• display_train_metrics –

• display_test_metrics –

Returns:

train_one_generation()

Standard one generation GA pipeline