def __init__(self, algorithm_configuration):
self.__algorithm_configuration = algorithm_configuration
self.__population_generator = PopulationGenerator(self.__algorithm_configuration)
self.__function = Function(self.__algorithm_configuration)
self.__elite_strategy = EliteStrategy(self.__algorithm_configuration)
self.__selection_service = SelectionService(self.__algorithm_configuration)
self.__cross_service = CrossService(self.__algorithm_configuration)
self.__mutation_service = MutationService(self.__algorithm_configuration)
self.__inversion_service = InversionService(self.__algorithm_configuration)
self.__chromosome_decoder = ChromosomeDecoder(self.__algorithm_configuration)
class TournamentSelection:
def __init__(self, algorithm_configuration):
self.__algorithm_configuration = algorithm_configuration
self.__group_members_amount = self.__algorithm_configuration.selection_amount
self.__function = Function(self.__algorithm_configuration)
def handle_selection(self, population):
np.random.shuffle(population)
groups = self.__split_into_groups(population,
self.__group_members_amount)
best_chromosomes = []
[
best_chromosomes.append(
self.__function.get_best_chromosome_for_population(group)[0])
for group in groups
]
return np.array(best_chromosomes)
@staticmethod
def __split_into_groups(population, members_amount):
groups = [
population[i * members_amount:(i + 1) * members_amount]
for i in range((len(population) + members_amount - 1) //
members_amount)
]
return groups
class BestStrategyBase:
def __init__(self, algorithm_configuration, number_of_best_chromosomes):
self.__number_of_best_chromosomes = number_of_best_chromosomes
self.__algorithm_configuration = algorithm_configuration
self.__function = Function(self.__algorithm_configuration)
def get_best_chromosomes(self, population):
if self.__number_of_best_chromosomes is 0:
return [], []
evaluated_population = self.__function.evaluate_population(population)
sorted_evaluated_population = evaluated_population.argsort()
optimization_evaluated_population = \
self.__get_sorted_evaluated_population_for_optimization(sorted_evaluated_population)
best_indexes = optimization_evaluated_population[-self.__number_of_best_chromosomes:]
best_chromosomes = np.take(population, best_indexes, 0)
new_population_to_evaluate = np.delete(population, best_indexes, 0)
return best_chromosomes, new_population_to_evaluate
def __get_sorted_evaluated_population_for_optimization(self, evaluated_population):
if self.__algorithm_configuration.is_maximization:
return evaluated_population
return evaluated_population[::-1]
class RouletteWheelSelection:
def __init__(self, algorithm_configuration):
self.__algorithm_configuration = algorithm_configuration
self.__function = Function(self.__algorithm_configuration)
self.__is_max_opt = self.__algorithm_configuration.is_maximization
def get_population(self, old_pop):
old_pop = np.sort(old_pop)
eval_pop = self.__function.evaluate_population(old_pop)
if not self.__is_max_opt:
eval_pop = 1/eval_pop
corrected_eval_pop = eval_pop + abs(eval_pop.min())
wheel = np.cumsum(corrected_eval_pop)
new_pop = []
for el in np.random.rand(self.__algorithm_configuration.selection_amount):
new_pop.append(old_pop[(wheel > el * wheel.max())][0])
return np.array(new_pop)
def __init__(self, algorithm_configuration):
self.__algorithm_configuration = algorithm_configuration
self.__group_members_amount = self.__algorithm_configuration.selection_amount
self.__function = Function(self.__algorithm_configuration)
def __init__(self, algorithm_configuration, number_of_best_chromosomes):
self.__number_of_best_chromosomes = number_of_best_chromosomes
self.__algorithm_configuration = algorithm_configuration
self.__function = Function(self.__algorithm_configuration)
from libs.config.algorithm_configuration_provider import AlgorithmConfigurationProvider from libs.generator.population_generator import PopulationGenerator from libs.algorithm.function import Function from libs.algorithm.genetic_algorithm import GeneticAlgorithm from libs.plot.plot_drawer import PlotDrawer algorithm_configuration = AlgorithmConfigurationProvider(2, 3, 3, 8, 10, True) population_generator = PopulationGenerator(algorithm_configuration) population = population_generator.generate_population() #print(population) # #print(chromosome_decoder.decode_chromosome(population[0])) #print(chromosome_decoder.decode_chromosome(population[1])) #print(chromosome_decoder.decode_chromosome(population[2])) function = Function(algorithm_configuration) #print(function.evaluate_population(population)) #print(function.get_best_chromosome_for_population(population)) #elite_strategy = EliteStrategy(algorithm_configuration, 1) #print(elite_strategy.get_best_chromosomes(population)) #best_strategy = BestStrategy(algorithm_configuration, 10) #print(best_strategy.get_best_chromosomes(population)) #tournament_selection = TournamentSelection(algorithm_configuration, 2) #print(tournament_selection.handle_selection(population)) genetic_algorithm = GeneticAlgorithm(algorithm_configuration) #print(genetic_algorithm.evolve())
def __init__(self, algorithm_configuration):
self.__algorithm_configuration = algorithm_configuration
self.__function = Function(self.__algorithm_configuration)
self.__is_max_opt = self.__algorithm_configuration.is_maximization
class GeneticAlgorithm:
def __init__(self, algorithm_configuration):
self.__algorithm_configuration = algorithm_configuration
self.__population_generator = PopulationGenerator(self.__algorithm_configuration)
self.__function = Function(self.__algorithm_configuration)
self.__elite_strategy = EliteStrategy(self.__algorithm_configuration)
self.__selection_service = SelectionService(self.__algorithm_configuration)
self.__cross_service = CrossService(self.__algorithm_configuration)
self.__mutation_service = MutationService(self.__algorithm_configuration)
self.__inversion_service = InversionService(self.__algorithm_configuration)
self.__chromosome_decoder = ChromosomeDecoder(self.__algorithm_configuration)
def evolve(self):
population = self.__population_generator.generate_population()
solution_best_value = self.__initialize_solution_best_value()
solution_best_chromosome = []
list_best = []
list_mean = []
list_std = []
start = time.time()
for i in range(self.__algorithm_configuration.epochs_number):
best_chromosomes, new_population_to_evaluate = self.__elite_strategy.get_best_chromosomes(population)
population = self.__selection_service.handle_selection(new_population_to_evaluate)
population = self.__cross_service.handle_cross(population)
population = self.__mutation_service.handle_mut(population)
population = self.__inversion_service.handle_inv(population)
population = np.concatenate((population, best_chromosomes), axis=0)
evaluated_population = self.__function.evaluate_population(population)
current_best_chromosome, current_best_chromosome_function_value = \
self.__function.get_best_chromosome_for_population(population)
solution_best_chromosome, solution_best_value = \
self.__update_solution_best_value(solution_best_value, current_best_chromosome_function_value,
solution_best_chromosome, current_best_chromosome)
list_best.append(current_best_chromosome_function_value)
list_mean.append(np.mean(evaluated_population))
list_std.append(np.std(evaluated_population))
end = time.time()
elapsed_time = end - start
decoded_best_chromosome = self.__chromosome_decoder.decode_chromosome(solution_best_chromosome)
return decoded_best_chromosome, solution_best_value, list_best, list_mean, list_std, elapsed_time
def __update_solution_best_value(self, solution_best_value, current_best_chromosome_function_value,
solution_best_chromosome, current_best_chromosome):
if self.__algorithm_configuration.is_maximization:
if solution_best_value < current_best_chromosome_function_value:
solution_best_chromosome = current_best_chromosome
return solution_best_chromosome, current_best_chromosome_function_value
else:
if solution_best_value > current_best_chromosome_function_value:
solution_best_chromosome = current_best_chromosome
solution_best_value = current_best_chromosome_function_value
return solution_best_chromosome, solution_best_value
def __initialize_solution_best_value(self):
if self.__algorithm_configuration.is_maximization:
return -float("inf")
return float("inf")