def evolve_best(individuals, fitness, n_parents):
"""
Breed only the best individuals
:param individuals: List of floats
:param fitness: List of floats
:param n_parents: Integer
:return: List
"""
# -- Select parents --
parents = selection.sel_best(
fitness, n_parents) # Indexes of best individuals in descending order
# -- Produce children --
children = breed.breed_uniform(individuals,
parents,
n_children=len(individuals),
co_prob=0.5)
# -- Mutate children --
for child in children:
mutation.mut_gauss(child, 0.01)
new_individuals = list(children)
# -- Elitism --
best_id = selection.sel_best(fitness, round(0.05 * len(fitness)))
for i, best in enumerate(best_id):
new_individuals[i] = list(individuals[best])
return new_individuals
def evolve_roulette(individuals, fitness, tournaments):
"""
Breed a new population using roulette selection.
:param individuals: List of floats
:param fitness: List of floats
:param tournaments: Integer
:return: List
"""
# -- Select parents --
parents = selection.sel_roulette(fitness, tournaments)
# -- Perform Crossover --
children = breed.breed_uniform(individuals,
parents,
n_children=len(individuals),
co_prob=0.5)
# ------- Mutate children -------
for child in children:
mutation.mut_gauss(child, 0.01)
new_individuals = list(children)
# -- Elitism --
best_id = selection.sel_best(fitness, round(0.05 * len(fitness)))
for i, best in enumerate(best_id):
new_individuals[i] = list(individuals[best])
return new_individuals
def evolve_tournament(individuals, fitness, tournaments, tour_size):
"""
Perform tournament selection and mutate children. Replace non-parents with children
:param individuals: List of floats
:param fitness: List of floats
:param tournaments: Integer
:param tour_size: Integer
:return: List
"""
# ------- Tournament selection --------
parents = selection.sel_tournament(fitness,
tournaments,
tour_size,
replace=False)
# -- Perform Crossover --
children = breed.breed_uniform(individuals,
parents,
n_children=len(individuals),
co_prob=0.5)
# ------- Mutate children -------
for child in children:
mutation.mut_gauss(child, 0.01)
new_individuals = list(children)
# -- Elitism --
best_id = selection.sel_best(fitness, round(0.05 * len(fitness)))
for i, best in enumerate(best_id):
new_individuals[i] = list(individuals[best])
return new_individuals
def evolve_breed_roulette(individuals, fitness, tournaments):
"""
Breed a new population using breed_roulette.
**
This method does - NOT - use roulette selection to pick parents from the population.
It uses roulette selection when picking which parents should breed.
See 'evolve_roulette' for using roulette selection.
**
:param individuals: List of floats
:param fitness: List of floats
:param tournaments: Integer
:return: List
"""
# -- Select parents --
parents = selection.sel_random(list(range(len(fitness))), tournaments)
parents_fitness = [fitness[val] for val in parents]
# -- Perform Crossover --
children = breed.breed_roulette(individuals,
parents,
parents_fitness,
n_children=len(individuals),
co_prob=0.5)
# ------- Mutate children -------
for child in children:
mutation.mut_gauss(child, 0.01)
new_individuals = list(children)
# -- Elitism --
best_id = selection.sel_best(fitness, round(0.05 * len(fitness)))
for i, best in enumerate(best_id):
new_individuals[i] = list(individuals[best])
return new_individuals