def test_apply_global_crossover():
np.random.seed(42)
size = 1000
ind_size = 100
operator = pg.intermediary_crossover
pop = pg.make_uniform_population(size, ind_size)
original_pop = pop.clone()
pop = pg.apply_global_crossover(pop, operator)
for i in range(pop.size):
assert pop.individuals[i].run_eval is True
def test_apply_global_mutation():
np.random.seed(42)
size = 30
ind_size = 100
operator = pg.uncorrelated_one_step_mutation
pool_size = 100
pop = pg.make_uniform_population(size, ind_size)
original_pop = pop.clone()
pop = pg.apply_global_mutation(pop, pool_size, operator)
assert pop.size == pool_size
for i in range(pop.size):
assert pop.individuals[i].run_eval is True
def test_m_plus_lambda():
np.random.seed(42)
sphere_model_size = 30
fitness_fn = lambda x: pg.sphere_model(x[:sphere_model_size])
size = 10
pop1 = pg.make_uniform_population(size, sphere_model_size)
pop1 = pg.evaluate_population(pop1, fitness_fn)
pop1_clone = pop1.clone()
pop2 = pg.make_uniform_population(size, sphere_model_size)
pop2 = pg.evaluate_population(pop2, fitness_fn)
pop1 = pg.mu_plus_lambda_replacement(pop1, pop2)
f_p1c = np.array(
[pop1_clone.individuals[i].fitness.value for i in range(size)])
f_p2 = np.array([pop2.individuals[i].fitness.value for i in range(size)])
f_p1 = np.array([pop1.individuals[i].fitness.value for i in range(size)])
res = np.array([
7.76332642, 8.08389311, 8.42498289, 8.77341594, 8.9523664, 9.45857558,
9.49688766, 9.52769953, 9.7264757, 9.83516821
])
assert np.allclose(f_p1, res) is True
assert np.allclose(f_p1, f_p1c) is False
assert f_p1.any() == f_p2.any()
def test_make_uniform_population():
size = 1000
ind_size = 20
low = 0.0
high = 1.0
pop = pg.make_uniform_population(size, ind_size, low=low, high=high)
assert type(pop) is pg.Population
assert pop.size == size
assert pop.individuals.size == size
for i in range(size):
assert type(pop.individuals[i]) is pg.Individual
assert pop.individuals[i].genotype.size == ind_size
for g in range(ind_size):
assert pop.individuals[i].genotype[g] >= low and pop.individuals[
i].genotype[g] <= high
def test_m_comma_lambda():
np.random.seed(42)
sphere_model_size = 30
fitness_fn = lambda x: pg.sphere_model(x[:sphere_model_size])
size = 10
pop1 = pg.make_uniform_population(size, sphere_model_size)
pop1 = pg.evaluate_population(pop1, fitness_fn)
pop1_clone = pop1.clone()
pop2 = pg.make_uniform_population(size, sphere_model_size)
pop2 = pg.evaluate_population(pop2, fitness_fn)
pop1 = pg.mu_comma_lambda_replacement(pop1, pop2)
f_p1c = np.array(
[pop1_clone.individuals[i].fitness.value for i in range(size)])
f_p2 = np.array([pop2.individuals[i].fitness.value for i in range(size)])
f_p1 = np.array([pop1.individuals[i].fitness.value for i in range(size)])
res = np.array([
7.7633264242708808, 8.4249828886076532, 9.4585755842303119,
9.8351682100669695, 10.109666113362849, 10.120464613773503,
11.148338435537882, 12.62638917659477, 12.68750163403401,
13.486951247705489
])
assert np.allclose(f_p1, res) is True
assert np.allclose(f_p1, f_p1c) is False
assert f_p1.any() == f_p2.any()