evolutionary_optimization.evolutionary_algorithm package

Submodules

evolutionary_optimization.evolutionary_algorithm.ea_data_model module

class PerformancePlotting(fitness_over_time, phenotype_over_time, genotype_over_time)

Bases: object

Dataclass storing information about the algorithm’s state overtime.

__init__(fitness_over_time, phenotype_over_time, genotype_over_time)

fitness_over_time: List[Union[float, int]]

genotype_over_time: List[Union[float, int, List[Union[float, int]]]]

phenotype_over_time: List[Union[float, int]]

evolutionary_optimization.evolutionary_algorithm.ea_utils module

class CreateGif2D(animation_data_x, animation_data_y, static_plot_data)

Bases: object

__init__(animation_data_x, animation_data_y, static_plot_data)

Initialise CreateGif2D class.

This class creates a gif of a two-dimensional phenotype.

Parameters

• animation_data_x (ndarray) – genotype value to be lapsed in the gif.

• animation_data_y (ndarray) – phenotype value to be lapsed in the gif.

• static_plot_data (PlottingData) – data to be used in the background of the gif e.g. the phenotype function.

animate(i)

Function called at each frame, plots the best phenotype / genotype pair.

generate_animation()

Generates gif animation.

plot_background_for_frame()

Initialise the background for the gif.

class CreateGif3D(animation_data_x, animation_data_y, animation_data_z, static_plot_data)

Bases: object

__init__(animation_data_x, animation_data_y, animation_data_z, static_plot_data)

animate(i)

generate_animation()

evolutionary_optimization.evolutionary_algorithm.evolution module

class Evolution(phenotype, fitness_function=<evolutionary_optimization.fitness_functions.implemented_fitness_functions.MaximizeFitnessFunction object>, number_of_individuals=100, number_of_generations=20, ratio_of_elite_individuals=0.1)

Bases: object

__init__(phenotype, fitness_function=<evolutionary_optimization.fitness_functions.implemented_fitness_functions.MaximizeFitnessFunction object>, number_of_individuals=100, number_of_generations=20, ratio_of_elite_individuals=0.1)

Initialise Evolution class.

The Evolution class performs evolutionary optimization of a function (a phenotype). It contains a population of individuals that are evaluated at every iteration (generation) of the algorithm.

Parameters

• phenotype (AbstractPhenotype) – a phenotype instance with the desired genotype.

• fitness_function (AbstractFitnessFunction) – desired fitness function from interface.

• number_of_individuals (int) – number of phenotype instances to be used within the population.

• number_of_generations (int) – number of times the algorithm will be run.

• ratio_of_elite_individuals (float) – proportion of best scoring phenotypes in population that will be kept to the next generation.

create_gif(function_data)

Create gif animation of the best phenotype/genotype pair over time.

evolve()

Perform evolutionary optimisation.

This function performs the evolutionary optimisation. Over number_of_generations it evaluates the population, updates the population (with crossover and/ or mutation as initialised). It then records the best fitness score at each generation.

plot_fitness_score_over_time()

Plot score of the best individual at each generation.

plot_phenotype_function_and_best_individuals(function_data)

Plot phenotype function and best individual phenotype values.

record_performance()

In place addition of fitness score, phenotype and genotype values of the current best individual.

This function performs in place addition of the current best fitness score, phenotype value and genotype value to the performance_over_time attribute. You can visualise fitness over time using the plot_fitness_score_over_time function, and you can visualise phenotype and genotype values over time using plot_phenotype_function_and_best_individuals function.

evolutionary_optimization.evolutionary_algorithm.population module

class Population(number_of_individuals, phenotype, ratio_of_elite_individuals)

Bases: object

__init__(number_of_individuals, phenotype, ratio_of_elite_individuals)

Create and store phenotypes used in the Evolution object.

The Population object is used by the Evolution object to create, store, evaluate and update phenotypes.

Parameters

• number_of_individuals (int) – number of phenotype instances i.e. individuals in the desired population.

• phenotype (AbstractPhenotype) – a phenotype instance with the desired genotype.

• ratio_of_elite_individuals (float) – proportion of best scoring phenotypes in population that will be kept to the next generation.

create_list_of_new_individuals(n_new_individuals)

Create a list of Individual instances.

Parameters

n_new_individuals (int) – number of desired individuals in list.

Return type

List[AbstractPhenotype]

Returns

List of random Phenotype instances of length n_new_individuals.

static crossover_for_population_segment(list_of_parents)

Perform crossover for a list of phenotypes.

This method creates a new list of phenotypes (children) based on the parents’ genotypes by calling the phenotype’s crossover method.

Parameters

list_of_parents (List[AbstractPhenotype]) – list of phenotypes which should be used to generate offspring.

Return type

List[AbstractPhenotype]

Returns

List of new Phenotype objects created from their parents’ genotypes.

evaluate_population(fitness_function)

Find best_individual in population by calculating fitness scores for all individuals.

For each individual in the population calculates the fitness score and stores the best individual in the population.best_individual attribute.

Parameters

fitness_function (AbstractFitnessFunction) – fitness function used to evaluate the phenotype.

sort_phenotypes_by_fitness_score(fitness_function)

Sort list of AbstractPhenotype by descending fitness score.

Return type

List[AbstractPhenotype]

split_elite_individuals(fitness_function)

Split list of individuals into elite and non-elite individuals.

The function will create two lists to separate out elite individuals i.e. ones with the highest fitness scores, from those with lower fitness scores. Top 10% of individuals (or at least 1) will be kept from a list. The rest will be assigned to a separate list, which will be updated using crossover and/or mutation.

Return type

Tuple[List[AbstractPhenotype], List[AbstractPhenotype]]

Returns

Tuple of List[Phenotype] and List[Phenotype] representing separate groups of

elite and non_elite individuals.

update_population(fitness_function)

Update population attribute following evaluation.

Once all individuals in the population have been evaluated, the top individuals are kept (elitism), the remaining individuals are updated by a combination of mutation and/or crossover if these were defined in the Evolution instance. If neither was selected, the non-elite individuals will be replaced by randomly generated individuals.

Parameters

fitness_function (AbstractFitnessFunction) – fitness function used to evaluate the phenotype.

get_score_for_sorting(phenotype_and_fitness_score_tuple)

Key for sorted function used in split_elite_individuals object.

Provides a key for sorting individuals by returning the phenotype’s fitness score from a tuple of phenotype and fitness score. This is used with python’s sorted function.

Parameters

phenotype_and_fitness_score_tuple (Tuple[AbstractPhenotype, int]) – tuple of phenotype instance and its fitness score.

Return type

Union[float, int]

Returns

Float or int equal to a phenotype’s fitness score.

Module contents