def breed(self, generation):
splicer = GeneSplicer()
children = []
for count in range(0, int(self.expected_children)):
if count == 0:
mum = copy.deepcopy(self.member_organisms[count])
else:
mum = copy.deepcopy(self.member_organisms[random.randint(0, len(self.member_organisms) - 1)])
if random.random() < MATE_PROBABILITY:
dad = copy.deepcopy(self.member_organisms[random.randint(0, len(self.member_organisms) - 1)])
junior = splicer.mate(mum, dad)
junior["reproduction"] = "Mate"
else:
junior = mum
splicer.mutate(junior)
junior["reproduction"] = "Mutate"
junior["parent"] = mum["id"]
junior["id"] = ControlPanel.next_organism_number()
junior["generation"] = generation
junior["species"] = self.id
children.append(junior)
mutants = []
for child in children:
if splicer.compatibility_scan(self.representative, child) < COMPATIBILITY_THRESHOLD:
self.member_organisms.append(child)
else:
mutants.append(child)
return mutants
def new_neuron(self):
next_innovation_number = ControlPanel.next_innovation_number()
return {
"innovation_number": next_innovation_number,
"label": "H%s" % next_innovation_number,
"type": HIDDEN
}
def __init__(self, representative):
self.id = ControlPanel.next_species_id()
self.representative = representative
self.member_organisms = [representative]
self.last_improvement = 0
self.max_fitness = 0
self.age = 0
self.expected_children = 0
def new_synapse(self, axon, dendrite, weight):
next_innovation_number = ControlPanel.next_innovation_number()
return {
"innovation_number": next_innovation_number,
"axon": axon,
"dendrite": dendrite,
"weight": weight
}
def __init__(self, representative):
self.id = ControlPanel.next_species_id()
self.representative = representative
self.member_organisms = [representative]
self.last_improvement = 0
self.max_fitness = 0
self.age = 0
self.expected_children = 0
def new_synapse(self, axon, dendrite, weight):
next_innovation_number = ControlPanel.next_innovation_number()
return {
"innovation_number": next_innovation_number,
"axon": axon,
"dendrite": dendrite,
"weight": weight
}
def mate(self, mum, dad):
mum_neurons = mum["neurons"]
dad_neurons = dad["neurons"]
mum_synapses = mum["synapses"]
dad_synapses = dad["synapses"]
better_fitness = mum if mum["fitness"] > dad["fitness"] else dad
junior_synapses = []
junior_neurons = {mum_neuron["label"]: mum_neuron for mum_neuron in mum_neurons if
mum_neuron["label"][0] != "H"}
g1 = 0
g2 = 0
while g1 < len(mum_synapses) and g2 < len(dad_synapses):
if g1 >= len(dad_synapses):
if better_fitness is mum:
junior_synapses.append(mum_synapses[g1])
self.mum_neurons(g1, junior_neurons, mum_neurons, mum_synapses[g1])
g1 += 1
elif g2 >= len(dad_synapses):
if better_fitness is not mum:
junior_synapses.append(dad_synapses[g2])
self.add_dad_neurons(g2, dad_neurons, junior_neurons, dad_synapses[g2])
g2 += 1
else:
innov1 = mum_synapses[g1]["innovation_number"]
innov2 = dad_synapses[g2]["innovation_number"]
if innov1 == innov2:
if rand_pos_neg() == 1:
junior_synapses.append(mum_synapses[g1])
self.mum_neurons(g1, junior_neurons, mum_neurons, mum_synapses[g1])
else:
junior_synapses.append(dad_synapses[g2])
self.add_dad_neurons(g2, dad_neurons, junior_neurons, dad_synapses[g2])
g1 += 1
g2 += 1
elif innov1 < innov2:
if better_fitness is mum:
junior_synapses.append(mum_synapses[g1])
self.mum_neurons(g1, junior_neurons, mum_neurons, mum_synapses[g1])
g1 += 1
else:
if better_fitness is not mum:
junior_synapses.append(dad_synapses[g1])
self.add_dad_neurons(g2, dad_neurons, junior_neurons, dad_synapses[g2])
g2 += 1
junior = {"id": ControlPanel.next_organism_number(), "fitness": mum["fitness"], "neurons": [], "synapses": []}
connections = set()
for i, synapse in enumerate(junior_synapses):
connection = synapse["axon"] + synapse["dendrite"]
if connection not in connections:
connections.add(connection)
junior["synapses"].append(synapse)
junior["neurons"].extend(junior_neurons.values())
return junior
def breed(self, generation):
splicer = GeneSplicer()
children = []
for count in range(0, int(self.expected_children)):
if count == 0:
mum = copy.deepcopy(self.member_organisms[count])
else:
mum = copy.deepcopy(self.member_organisms[random.randint(
0,
len(self.member_organisms) - 1)])
if random.random() < MATE_PROBABILITY:
dad = copy.deepcopy(self.member_organisms[random.randint(
0,
len(self.member_organisms) - 1)])
junior = splicer.mate(mum, dad)
junior["reproduction"] = "Mate"
else:
junior = mum
splicer.mutate(junior)
junior["reproduction"] = "Mutate"
junior["parent"] = mum["id"]
junior["id"] = ControlPanel.next_organism_number()
junior["generation"] = generation
junior["species"] = self.id
children.append(junior)
mutants = []
for child in children:
if splicer.compatibility_scan(self.representative,
child) < COMPATIBILITY_THRESHOLD:
self.member_organisms.append(child)
else:
mutants.append(child)
return mutants
def mate(self, mum, dad):
mum_neurons = mum["neurons"]
dad_neurons = dad["neurons"]
mum_synapses = mum["synapses"]
dad_synapses = dad["synapses"]
better_fitness = mum if mum["fitness"] > dad["fitness"] else dad
junior_synapses = []
junior_neurons = {
mum_neuron["label"]: mum_neuron
for mum_neuron in mum_neurons if mum_neuron["label"][0] != "H"
}
g1 = 0
g2 = 0
while g1 < len(mum_synapses) and g2 < len(dad_synapses):
if g1 >= len(dad_synapses):
if better_fitness is mum:
junior_synapses.append(mum_synapses[g1])
self.mum_neurons(g1, junior_neurons, mum_neurons,
mum_synapses[g1])
g1 += 1
elif g2 >= len(dad_synapses):
if better_fitness is not mum:
junior_synapses.append(dad_synapses[g2])
self.add_dad_neurons(g2, dad_neurons, junior_neurons,
dad_synapses[g2])
g2 += 1
else:
innov1 = mum_synapses[g1]["innovation_number"]
innov2 = dad_synapses[g2]["innovation_number"]
if innov1 == innov2:
if rand_pos_neg() == 1:
junior_synapses.append(mum_synapses[g1])
self.mum_neurons(g1, junior_neurons, mum_neurons,
mum_synapses[g1])
else:
junior_synapses.append(dad_synapses[g2])
self.add_dad_neurons(g2, dad_neurons, junior_neurons,
dad_synapses[g2])
g1 += 1
g2 += 1
elif innov1 < innov2:
if better_fitness is mum:
junior_synapses.append(mum_synapses[g1])
self.mum_neurons(g1, junior_neurons, mum_neurons,
mum_synapses[g1])
g1 += 1
else:
if better_fitness is not mum:
junior_synapses.append(dad_synapses[g1])
self.add_dad_neurons(g2, dad_neurons, junior_neurons,
dad_synapses[g2])
g2 += 1
junior = {
"id": ControlPanel.next_organism_number(),
"fitness": mum["fitness"],
"neurons": [],
"synapses": []
}
connections = set()
for i, synapse in enumerate(junior_synapses):
connection = synapse["axon"] + synapse["dendrite"]
if connection not in connections:
connections.add(connection)
junior["synapses"].append(synapse)
junior["neurons"].extend(junior_neurons.values())
return junior
def new_neuron(self):
next_innovation_number = ControlPanel.next_innovation_number()
return {"innovation_number": next_innovation_number,
"label": "H%s" % next_innovation_number,
"type": HIDDEN
}
