def calculate_standard_deviation(self, graph, population):
import math
fitness_list = self.evaluate_fitness(graph, population)
avg_score = np.mean(
[goalFunction(graph.applyMask(i)) for i in population])
variance = np.mean([math.pow(i - avg_score, 2) for i in fitness_list])
deviation = math.sqrt(variance)
return deviation, avg_score
def run(self, graph, population, params, num_parents, debug):
start = time.time()
population = [
self.generateRandomMask(graph) for i in range(params['population'])
]
t_condition, avg_score, deviation = self.ga.termination_method(
graph, population)
while (t_condition):
# Measuring the fitness of each chromosome in the population. And Updating the population
fitness_list, population = self.ga.succesion_method(
population, graph)
# Selecting best mates(Parents) and add them to population
parents = self.ga.selection(population, fitness_list, num_parents)
# Generate crossover
if (params['cross_prob'] > random.random()):
children = self.ga.cross_method(population, parents)
if (params['mutation_prob'] > random.random()):
# Generate Mutations
children_mutations = self.ga.mutation(population, children)
# Update Population and mesuare new fitness
fitness_list = self.ga.evaluate_fitness(graph, population)
# Update t_condition
t_condition, avg_score, deviation = self.ga.termination_method(
graph, population)
#Update score
graph_indx = np.argmin(fitness_list)
mask = population[graph_indx]
score = goalFunction(graph.applyMask(mask))
if (score < self.score):
self.history.append(score)
self.bestGraph = copy.deepcopy(graph.applyMask(mask))
self.score = goalFunction(self.bestGraph)
# Debug
if (debug):
self.debug.show_iterations(avg_score, deviation)
#self.debug.show_parent_children(parents, children, children_mutations, graph)
end = time.time()
self.time = end - start
def roulete(self, fitness_list, population, graph):
summation = np.sum(
[goalFunction(graph.applyMask(i)) for i in population])
u = random.random()
if (u < 0.3):
u = 0.3
cur_population = len(population)
new_population = int(cur_population * u)
#print("cur: {} new: {} u: {}".format(cur_population, new_population, u))
while (len(population) > new_population
and len(fitness_list) > new_population and new_population > 4):
del population[np.argmax(fitness_list)]
del fitness_list[np.argmax(fitness_list)]
return fitness_list, population
#endregion
def evaluate_fitness(self, graph, population):
return [goalFunction(graph.applyMask(i)) for i in population]
def show_parent_children(self, generations, parents, children,
children_mutations, graph):
child = goalFunction(graph.applyMask(children[0]))
print("Generation: {} Parent1: {} parent2: {} Children: {}".format(
generations, parents[0][0], parents[1][0], child))
