def test_mutate():
parent = ClassicGRN()
parent.random(5, 5, 5)
config.MUTATION_ADD_RATE = 1.0
child = evolution.mutate(parent)
assert parent.size() == child.size() - 1
assert parent.distance_to(child) < 1 / (child.size() + 2)
config.MUTATION_ADD_RATE = 0.0
config.MUTATION_DEL_RATE = 1.0
child = evolution.mutate(parent)
assert parent.size() == child.size() + 1
assert parent.distance_to(child) < 1 / (child.size() + 2)
config.MUTATION_DEL_RATE = 0.0
child = evolution.mutate(parent)
assert parent.size() == child.size()
assert parent.distance_to(child) < 1 / (child.size() + 2)
def test_evaluate():
ind = evolution.Individual(ClassicGRN())
problem = Random()
ind.grn.random(problem.nin, problem.nout, 1)
assert ind.fitness == 0.0
fitness = ind.get_fitness(problem)
assert ind.fitness == fitness
assert fitness != 0
def __init__(self, problem, new_grn_function=lambda: ClassicGRN(),
run_id=str(uuid4()), grn_dir='grns', log_dir='logs'):
pathlib.Path(grn_dir).mkdir(parents=True, exist_ok=True)
pathlib.Path(log_dir).mkdir(parents=True, exist_ok=True)
self.grn_file = os.path.join(grn_dir, 'grns_' + run_id + '.log')
self.log_file = os.path.join(log_dir, 'fits_' + run_id + '.log')
self.problem = problem
self.population = Population(new_grn_function, problem.nin,
problem.nout)
self.generation = 0
def test_thresholds():
problem = Random()
config.INITIALIZATION_DUPLICATION = 8
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
pop.evaluate(problem)
pop.speciation()
pop.adjust_thresholds()
assert np.any([sp.species_threshold for sp in pop.species] !=
config.START_SPECIES_THRESHOLD)
def test_best():
problem = Counter()
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
pop.evaluate(problem)
assert problem.count == config.POPULATION_SIZE
pop.speciation()
best_fit, best_ind = pop.get_best()
assert problem.count == config.POPULATION_SIZE
assert best_fit > 0
def test_mutate_remove():
parent = ClassicGRN()
parent.random(5, 5, 5)
child = evolution.mutate_remove(parent)
assert len(parent.inhibitors) == len(child.inhibitors) + 1
assert len(parent.enhancers) == len(child.enhancers) + 1
assert len(parent.identifiers) == len(child.identifiers) + 1
assert parent.size() == child.size() + 1
assert parent.distance_to(child) < 1 / (parent.size() + 2)
def test_offspring_count():
problem = Random()
config.INITIALIZATION_DUPLICATION = 8
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
pop.evaluate(problem)
pop.speciation()
pop.adjust_thresholds()
pop.set_offspring_count()
assert (np.sum([sp.num_offspring for sp in pop.species]) ==
config.POPULATION_SIZE)
def test_speciation():
problem = Counter()
config.INITIALIZATION_DUPLICATION = 8
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
pop.evaluate(problem)
pop.speciation()
assert len(pop.offspring) == 0
assert len(pop.species) > 0
print([len(sp.individuals) for sp in pop.species])
assert problem.count == config.POPULATION_SIZE
def test_fit_increase():
problem = Counter()
problem.cacheable = True
evo = Evolution(problem, lambda: ClassicGRN())
evo.step()
assert np.any([sp.sum_adjusted_fitness > 0.0
for sp in evo.population.species])
best_fit, best_ind = evo.population.get_best()
evo.run(10)
new_best_fit, new_best_ind = evo.population.get_best()
assert np.any([sp.sum_adjusted_fitness > 0.0
for sp in evo.population.species])
assert new_best_fit >= best_fit
def test_sum_adjusted_fitness():
species = evolution.Species()
assert species.sum_adjusted_fitness is None
problem = Random()
for i in range(10):
ind = evolution.Individual(ClassicGRN())
ind.grn.random(problem.nin, problem.nout, 1)
species.individuals += [ind]
assert species.get_adjusted_fitness(0.0, 1.0) == 0.0
assert species.sum_adjusted_fitness == 0.0
for ind in species.individuals:
ind.get_fitness(problem)
assert species.get_adjusted_fitness(0.0, 1.0) > 0.0
assert species.sum_adjusted_fitness > 0.0
def test_make_offspring():
problem = Counter()
config.INITIALIZATION_DUPLICATION = 8
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
pop.evaluate(problem)
pop.speciation()
pop.adjust_thresholds()
pop.set_offspring_count()
pop.make_offspring()
assert problem.count == config.POPULATION_SIZE
assert len(pop.offspring) == config.POPULATION_SIZE
assert np.any([ind.evaluated for ind in pop.offspring])
assert np.any([~ind.evaluated for ind in pop.offspring])
def test_mutate_modify():
parent = ClassicGRN()
parent.random(5, 5, 5)
child = evolution.mutate_modify(parent)
if parent.beta == child.beta and parent.delta == child.delta:
assert (np.sum(parent.inhibitors != child.inhibitors) +
np.sum(parent.enhancers != child.enhancers) +
np.sum(parent.identifiers != child.identifiers)) == 1
assert len(parent.inhibitors) == len(child.inhibitors)
assert len(parent.enhancers) == len(child.enhancers)
assert len(parent.identifiers) == len(child.identifiers)
assert parent.size() == child.size()
assert parent.distance_to(child) < 1 / (parent.size() + 2)
def test_representative_distances():
species = evolution.Species()
assert species.sum_adjusted_fitness is None
problem = Random()
rep = evolution.Individual(ClassicGRN())
rep.grn.random(problem.nin, problem.nout, 1)
species.representative = rep
for i in range(10):
ind = evolution.Individual(evolution.mutate(rep.grn))
species.individuals += [ind]
dists0 = species.get_representative_distances()
assert np.sum(dists0) > 0
rep.grn.random(problem.nin, problem.nout, 1)
dists1 = species.get_representative_distances()
assert np.sum(dists1) > 0
assert np.sum(dists1) > np.sum(dists0)
def test_noncacheable_counter():
problem = Counter()
problem.cacheable = False
evo = Evolution(problem, lambda: ClassicGRN())
evo.step()
assert np.any([sp.sum_adjusted_fitness > 0.0
for sp in evo.population.species])
assert problem.count == config.POPULATION_SIZE
evo.step()
assert np.any([sp.sum_adjusted_fitness > 0.0
for sp in evo.population.species])
assert problem.count == 2 * config.POPULATION_SIZE
evo.step()
assert np.any([sp.sum_adjusted_fitness > 0.0
for sp in evo.population.species])
assert problem.count == 3 * config.POPULATION_SIZE
def test_mutate_add():
parent = ClassicGRN()
parent.random(5, 5, 5)
child = evolution.mutate_add(parent)
assert np.all(parent.inhibitors == child.inhibitors[0:-1])
assert np.all(parent.enhancers == child.enhancers[0:-1])
assert np.all(parent.identifiers == child.identifiers[0:-1])
assert len(parent.inhibitors) == len(child.inhibitors) - 1
assert len(parent.enhancers) == len(child.enhancers) - 1
assert len(parent.identifiers) == len(child.identifiers) - 1
assert parent.size() == child.size() - 1
assert parent.distance_to(child) < 1 / (child.size() + 2)
def test_evaluation():
problem = Counter()
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
pop.evaluate(problem)
assert problem.count == config.POPULATION_SIZE
def test_eq_size_crossover():
parent1 = ClassicGRN()
parent2 = ClassicGRN()
parent1.random(5, 5, 5)
parent2.random(5, 5, 5)
child = evolution.crossover(parent1, parent2)
assert (child.beta == parent1.beta) or (child.beta == parent2.beta)
assert (child.delta == parent1.delta) or (child.delta == parent2.delta)
assert parent1.size() == child.size()
assert parent2.size() == child.size()
assert parent1.distance_to(child) < parent1.distance_to(parent2)
assert parent2.distance_to(child) < parent2.distance_to(parent1)
assert parent1.distance_to(child) > 0
assert parent2.distance_to(child) > 0
def test_init():
problem = Counter()
pop = evolution.Population(lambda: ClassicGRN(),
problem.nin, problem.nout)
assert len(pop.offspring) == config.POPULATION_SIZE
assert len(pop.species) == 0