def test_reproduce():
conf = Conf()
innovs = Innovations()
genome = Genome.minimal_fully_connected(0, (2, 3))
species = Species()
ranks = random.sample(xrange(20), 10)
for x in xrange(10):
new_genome = genome.duplicate(innovs.next_genome())
org = Organism(new_genome)
org.rank = ranks[x]
species.organisms.append(org)
species.offspring = 10
children = species.epoch(conf, innovs)
assert len(children) == 10
def test_mutate_neuron():
genome = Genome.minimal_fully_connected(0, (3, 2))
innovations = Innovations()
innovations.innov = max(map(lambda x: x.innov, genome.genes))
genome.mutate_neuron(innovations)
assert len(genome.genes) == 8
assert sum(genome.neurons) == 6
assert len(innovations.neuron_innov) == 1
def test_compatible_diff_struct():
conf = Conf()
innovations = Innovations()
g1 = Genome.minimal_fully_connected(0, (3, 2))
innovations.innov = max(map(lambda x: x.innov, g1.genes))
g2 = g1.duplicate(1)
g2.mutate_neuron(innovations)
assert not g1.compatible(conf, g2)
def test_crossover_unequal_dad_mut_mom_fit():
innovs = Innovations()
mom = Genome.minimal_fully_connected(0, (3, 2))
innovs.innov = max(map(lambda x: x.innov, mom.genes))
dad = mom.duplicate(1)
dad.mutate_neuron(innovs)
baby = mom.crossover(dad, 2.0, 1.0, innovs)
assert len(baby.genes) == 6
def test_mutate_gene_none_free():
genome = Genome.minimal_fully_connected(0, (3, 2))
innovations = Innovations()
innovations.innov = max(map(lambda x: x.innov, genome.genes))
old = random.choice
random.choice = mock.MagicMock(side_effect=[0, Genome.MAX_HIDDEN])
genome.mutate_gene(innovations)
random.choice = old
assert len(genome.genes) == 6
assert len(innovations.gene_innov) == 0
def test_mutate_gene():
genome = Genome.minimal_fully_connected(0, (3, 2))
genome.neurons[1] += 1
innovations = Innovations()
innovations.innov = max(map(lambda x: x.innov, genome.genes))
old = random.choice
random.choice = mock.MagicMock(side_effect=[0, 3])
genome.mutate_gene(innovations)
random.choice = old
assert len(genome.genes) == 7
assert len(innovations.gene_innov) == 1