def classical_genetic_algorithm():
POPULATION_SIZE = 140
CROSSOVER_PROBABILITY = .3
MUTATION_PROBABILITY = .9
MAX_GENERATIONS = 140
Individual.count = 0
first_population = [
generate_random(len(points)) for _ in range(POPULATION_SIZE)
]
best_ind = random.choice(first_population)
fit_avg = []
fit_best = []
generation_num = 0
population = first_population.copy()
while generation_num < MAX_GENERATIONS:
generation_num += 1
offspring = selection_rank_with_elite(population, elite_size=2)
crossed_offspring = crossover_operation(offspring, crossover,
CROSSOVER_PROBABILITY)
mutated_offspring = mutation_operation(crossed_offspring, mutate,
MUTATION_PROBABILITY)
population = mutated_offspring.copy()
best_ind, fit_avg, fit_best = stats(population, best_ind, fit_avg,
fit_best)
return best_ind.fitness, Individual.count
def adaptive_genetic_algorithm():
POPULATION_SIZE = 200
MIN_POPULATION_SIZE = 50
MAX_POPULATION_SIZE = 300
CROSSOVER_PROBABILITY = .5
MIN_CROSSOVER_PROBABILITY = .1
MUTATION_PROBABILITY = .5
MIN_MUTATION_PROBABILITY = .1
MAX_GENERATIONS = 10_000
MIN_GENERATIONS = 100
ELITE_SIZE = 1
Individual.count = 0
first_population = [
generate_random(len(points)) for _ in range(POPULATION_SIZE)
]
best_ind = random.choice(first_population)
fit_avg = []
fit_best = []
learn_trend = []
ev_avg = []
population_size = []
mutation_prob = []
crossover_prob = []
generation_num = 0
population = first_population.copy()
while generation_num < MIN_GENERATIONS or\
(generation_num < MAX_GENERATIONS and is_learning_positive(fit_best, 50, .001)):
generation_num += 1
offspring = selection_rank_with_elite(population,
elite_size=ELITE_SIZE)
crossed_offspring = crossover_operation(offspring, crossover,
CROSSOVER_PROBABILITY)
mutated_offspring = mutation_operation(crossed_offspring, mutate,
MUTATION_PROBABILITY)
population = mutated_offspring.copy()
best_ind, fit_avg, fit_best = stats(population, best_ind, fit_avg,
fit_best)
ev_avg.append(average(fit_avg, 10))
learn_rate = is_learning_positive(fit_avg, 10, .001)
learn_trend.append(learn_rate)
if not learn_rate:
MUTATION_PROBABILITY = min(MUTATION_PROBABILITY * 1.1, 1)
CROSSOVER_PROBABILITY = min(CROSSOVER_PROBABILITY * 1.1, 1)
if len(population) < MAX_POPULATION_SIZE:
population = population + [
generate_random(len(points)) for _ in range(2)
]
else:
MUTATION_PROBABILITY = max(MUTATION_PROBABILITY * .99,
MIN_MUTATION_PROBABILITY)
CROSSOVER_PROBABILITY = max(CROSSOVER_PROBABILITY * .99,
MIN_CROSSOVER_PROBABILITY)
if len(population) > MIN_POPULATION_SIZE:
worst_ind = min(population, key=lambda ind: ind.fitness)
population.remove(worst_ind)
population_size.append(len(population))
crossover_prob.append(CROSSOVER_PROBABILITY)
mutation_prob.append(MUTATION_PROBABILITY)
return best_ind.fitness, Individual.count
ELITE_SIZE = 2
first_population = [
generate_random(len(points)) for _ in range(POPULATION_SIZE)
]
best_ind = random.choice(first_population)
fit_avg = []
fit_best = []
impr_list = []
ev_avg = []
generation_num = 0
population = first_population.copy()
while generation_num < MAX_GENERATIONS:
generation_num += 1
offspring = selection_rank_with_elite(population, elite_size=ELITE_SIZE)
crossed_offspring = crossover_operation(offspring, crossover,
CROSSOVER_PROBABILITY)
mutated_offspring = mutation_operation(crossed_offspring, mutate,
MUTATION_PROBABILITY)
population = mutated_offspring.copy()
best_ind, fit_avg, fit_best = stats(population, best_ind, fit_avg,
fit_best)
ev_avg.append(average(fit_avg, 10))
impr_list.append(is_improvement_positive(fit_avg, 10, .001))
def add_degradation_areas(learn_trend):
for i in range(len(learn_trend)):
if not learn_trend[i]:
plt.axvspan(i, i + 1, color='red', alpha=0.1)
def selection(population):
return selection_rank_with_elite(population, elite_size=1)