def test_mutate_binomial_err1():
"""If we fail to provide either expected_num_mutations or a probability parameter,
an exception should occur when the operator is used."""
mutator = intrep_ops.mutate_binomial(std=1, bounds=[(0, 1), (0, 1)])
ind1 = Individual(np.array([0, 0]))
ind2 = Individual(np.array([1, 1]))
population = iter([ind1, ind2])
result = mutator(population)
with pytest.raises(ValueError):
# Pulse the iterator so mutation gets executed
result = list(result)
def test_mutate_binomial_dist():
"""When we apply binomial mutation repeatedly, the resulting distribution
of offspring should follow the expected theoretical distribution."""
N = 5000 # Number of mutantes to generate
binom_n = 10000 # "coin flips" parameter for the binomial
std = 2.5 # Standard deviation of the mutation distribution
# We'll set up our operator with infinite bounds, so we needn't worry about clipping
operator = intrep_ops.mutate_binomial(std=std, expected_num_mutations=2,
bounds=[(-float('inf'), float('inf')), (-float('inf'), float('inf'))])
# Any value could appear, but we'll focus on measuring just a few
# nearby values
genome = np.array([5, 10])
gene0_observed_dist = { '3': 0, '4': 0, '5': 0, '6': 0, '7':0 }
gene1_observed_dist = { '8': 0, '9': 0, '10': 0, '11': 0, '12': 0 }
# Count the observed mutations in N trials
for i in range(N):
population = iter([ Individual(genome.copy()) ])
mutated = next(operator(population))
gene0, gene1 = mutated.genome
gene0, gene1 = str(int(gene0)), str(int(gene1))
# Count the observed values of the first gene
if gene0 in gene0_observed_dist.keys():
gene0_observed_dist[gene0] += 1
# Count the observed values of the second gene
if gene1 in gene1_observed_dist.keys():
gene1_observed_dist[gene1] += 1
# Set up the expected distribution by using SciPy's binomial PMF function
binom_p = intrep_ops._binomial_p_from_std(binom_n, std)
binom = stats.binom(binom_n, binom_p)
mu = binom_n * binom_p # Mean of a binomial distribution is n*p
gene0_expected_dist = { k: int(N*binom.pmf(int(mu - (genome[0] - int(k))))) for k in gene0_observed_dist.keys() }
gene1_expected_dist = { k: int(N*binom.pmf(int(mu - (genome[1] - int(k))))) for k in gene1_observed_dist.keys() }
# Toss all the other values under one value
gene0_observed_dist['other'] = N - sum(gene0_observed_dist.values())
gene1_observed_dist['other'] = N - sum(gene1_observed_dist.values())
gene0_expected_dist['other'] = N - sum(gene0_expected_dist.values())
gene1_expected_dist['other'] = N - sum(gene1_expected_dist.values())
p = 0.01
assert(stat.stochastic_equals(gene0_expected_dist, gene0_observed_dist, p=p))
assert(stat.stochastic_equals(gene1_expected_dist, gene1_observed_dist, p=p))
def test_mutate_binomial_bounds():
"""If we apply a wide mutation distribution repeatedly, it should never stray
outside of the provided bounds.
This test runs the stochastic function repeatedly, but we don't mark it as a
stochastic test because it's and should never fail unless there is actually a
fault."""
operator = intrep_ops.mutate_binomial(std=20, bounds=[(0, 10), (2, 20)],
expected_num_mutations=1)
N = 100
for i in range(N):
population = iter([ Individual(np.array([5,10])) ])
mutated = next(operator(population))
assert(mutated.genome[0] >= 0)
assert(mutated.genome[0] <= 10)
assert(mutated.genome[1] >= 2)
assert(mutated.genome[1] <= 20)
problem=problem, # Fitness function
# Representation
representation=Representation(
# Initialize a population of integer-vector genomes
initialize=create_int_vector(bounds=[problem.bounds] * l)),
# Operator pipeline
pipeline=[
ops.tournament_selection(k=2),
ops.clone,
# Apply binomial mutation: this is a lot like
# additive Gaussian mutation, but adds an integer
# value to each gene
mutate_binomial(std=1.5,
bounds=[problem.bounds] * l,
expected_num_mutations=1),
ops.evaluate,
ops.pool(size=pop_size),
# Some visualization probes so we can watch what happens
probe.CartesianPhenotypePlotProbe(xlim=problem.bounds,
ylim=problem.bounds,
contours=problem),
probe.FitnessPlotProbe(),
probe.PopulationMetricsPlotProbe(
metrics=[probe.pairwise_squared_distance_metric],
title='Population Diversity'),
probe.FitnessStatsCSVProbe(stream=sys.stdout)
])