def _simulate_population(cls, i, seed_offset, population_size,
string_size):
seed(19520624 + 10 * i + seed_offset)
creator = Creator(population_size=population_size,
string_size=string_size)
population = creator.create_population()
evolved_population, evaluations = creator.evolve_population(population)
success = False
if cls._found_global_optimum(creator, evolved_population, string_size):
success = True
return success, evaluations
def find_minimum_population_size(cls,
string_size,
initial_population_size=5,
num_runs=5) -> int:
population_size = initial_population_size
lower_bound = 0
upper_bound = 81920
found_upper_bound = False
found_global_optimum_count = 0
while lower_bound != upper_bound or found_global_optimum_count != num_runs:
found_global_optimum_count = 0
creator = Creator(population_size=population_size,
string_size=string_size)
population = creator.create_population()
for _ in range(num_runs):
evolved_population = creator.evolve_population(population)
if cls._found_global_optimum(creator, evolved_population,
string_size):
found_global_optimum_count += 1
if found_global_optimum_count > 0: # Found global optimum.
upper_bound = population_size
if not found_upper_bound:
lower_bound = int(population_size / 2)
population_size = cls._midpoint(lower_bound, upper_bound)
found_upper_bound = True
else: # Failed to find global optimum 5/5 times
if lower_bound == upper_bound: # Lower and upper bound converged
# Increment population size until it succeeds 5/5 times
population_size += 2
lower_bound = population_size
upper_bound = population_size
elif found_upper_bound: # Found upper bound and failed.
# Increase lower bound and population size.
population_size = cls._midpoint(lower_bound, upper_bound)
lower_bound = population_size
if upper_bound - lower_bound == 2:
lower_bound += 2
else: # Failed without finding upper bound.
population_size *= 2 # Double population size
return upper_bound
