def select_elite(self, population):
scores_for_chromosomes = []
for i in range(0, len(population)):
chromosome = population[i]
result = brainfuck.evaluate(chromosome)
score = self.fitness_score(result, self.target)
if score == self.max_fitness_score:
current_time = time.time()
print("\nFOUND SOLUTION: " + chromosome + " for: " + repr(self.target) + " in: " + str(int((current_time-self.start_time)/60)) + " minutes")
self.best_fitness_scores.append(self.max_fitness_score)
self.update_fitness_plot()
exit()
scores_for_chromosomes.append((chromosome, score))
scores_for_chromosomes.sort(key=lambda x: x[1])
scores = [x[1] for x in scores_for_chromosomes]
print("population: " + "(min: " + str(min(scores)) + ", avg: " + str(mean(scores)) + ", max: " + str(max(scores)) + ")")
top_performers = scores_for_chromosomes[-TOP_PERFORMERS_COUNT:]
top_scores = [x[1] for x in top_performers]
print("elite " + str(round(1.0-SELECTION_RATE, 2)) + ": " + "(min: " + str(min(top_scores)) + ", avg: " + str(mean(top_scores)) + ", max: " + str(max(top_scores)) + ")")
chromosome = top_performers[-1][0]
result = brainfuck.evaluate(chromosome)
best_fitness_score = self.fitness_score(result, self.target)
print("best: " + chromosome + ", result: " + repr(result) + ", score: " + str(best_fitness_score) + "/" + str(self.max_fitness_score))
self.best_fitness_scores.append(best_fitness_score)
self.update_fitness_plot()
return top_performers
def genetic_evolution(self):
self.population = self.generate_population(self.population)
while True:
print("\ngeneration: " + str(self.generation) + ", population: " + str(len(self.population)) + ", mutation_rate: " + str(MUTATION_RATE))
# 1. Selection
elite = self.select_elite(self.population)
# 2. Crossover (Roulette selection)
pairs = self.generate_pairs(elite)
selected_offsprings = []
for pair in pairs:
offsprings = self.crossover(pair[0][0], pair[1][0])
selected_offsprings.append(offsprings[random.randint(0, 1)])
# 3. Mutation
mutated_population = self.mutation(selected_offsprings)
# 4. Validation (We don't want syntactically incorrect programs)
valid_population = []
for chromosome in mutated_population:
if brainfuck.evaluate(chromosome) is not None:
valid_population.append(chromosome)
print("propagated to next generation: " + str(len(valid_population)))
self.population = self.generate_population(valid_population)
self.generation += 1
def mutation(self, selected_offsprings):
offsprings = []
for offspring in selected_offsprings:
valid = False
mutation_attempts = 0
offspring_mutation = copy.deepcopy(offspring)
while not valid and mutation_attempts < MAX_MUTATION_ATTEMPTS:
for i in range(0, len(offspring_mutation)):
if np.random.choice([True, False],
p=[MUTATION_RATE, 1 - MUTATION_RATE]):
action_type = random.randint(0, 2)
if action_type == 0 and len(
offspring_mutation
) < PROGRAM_LENGTH_UPPER_BOUND:
# Inserting random value at index
offspring_mutation = offspring_mutation[:i] + random.choice(
AVAILABLE_OPS) + offspring_mutation[i:]
elif action_type == 1 and len(
offspring_mutation
) > PROGRAM_LENGTH_LOWER_BOUND:
# Removing value at index
offspring_mutation = offspring_mutation[:
i] + offspring_mutation[
i + 1:]
else:
# Setting random value at index
offspring_mutation = self.set_value_at_index(
offspring_mutation,
random.choice(AVAILABLE_OPS), i)
if brainfuck.evaluate(offspring_mutation) is not None:
valid = True
offsprings.append(offspring_mutation)
mutation_attempts += 1
return offsprings
def generate_population(self, population): while len(population) < POPULATION: length = random.randint(PROGRAM_LENGTH_LOWER_BOUND, PROGRAM_LENGTH_UPPER_BOUND) chromosome = "" for i in range(0, length): chromosome += random.choice(AVAILABLE_OPS) if brainfuck.evaluate(chromosome) is not None: # We don't want programs that are syntactically incorrect print(chromosome) exit() population.append(chromosome) return population