def two_point(p1, p2):
# one or two point crossover
point1 = rng.randrange(1, len(p1.measures))
point2 = rng.randrange(1, len(p1.measures))
while point1 == point2:
point1 = rng.randrange(1, len(p1.measures))
point2 = rng.randrange(1, len(p1.measures))
if point1 > point2:
temp = point2
point2 = point1
point1 = temp
c1 = Individual([], None)
c2 = Individual([], None)
p1_copy = copy.deepcopy(p1)
p2_copy = copy.deepcopy(p2)
for i in range(0, point1):
c1.measures.append(p1_copy.measures[i])
c2.measures.append(p2_copy.measures[i])
for i in range(point1, point2):
c1.measures.append(p2_copy.measures[i])
c2.measures.append(p1_copy.measures[i])
for i in range(point2, len(p1_copy.measures)):
c1.measures.append(p1_copy.measures[i])
c2.measures.append(p2_copy.measures[i])
if c1.measures[0] is p1.measures[0]:
print("Child measure has reference to parent measure")
return c1, c2
def create(self, genotype=None):
if genotype is None:
genotype = []
for i in range(0, config.length):
genotype += self.to_binary(random.randint(0, config.alphabet-1), math.ceil(math.log2(config.alphabet)))
return Individual(config.phenotype_convertor, config.fitness_evaluator, config.mutation_strategy, genotype,
config.crossover_strategy)
def create(self, genotype=None):
gene_length = config.gene_length
if genotype is None:
genotype = numpy.random.choice([0, 1],
size=(gene_length, )).tolist()
return Individual(config.phenotype_convertor, config.fitness_evaluator,
config.mutation_strategy, genotype,
config.crossover_strategy)
def uniform(p1, p2):
c1 = Individual([], None)
c2 = Individual([], None)
p1_copy = copy.deepcopy(p1)
p2_copy = copy.deepcopy(p2)
for i in range(len(p1_copy.measures)):
p = rng.choice([True, False])
if p:
c1.measures.append(p2_copy.measures[i])
c2.measures.append(p1_copy.measures[i])
else:
c1.measures.append(p1_copy.measures[i])
c2.measures.append(p2_copy.measures[i])
if c1.measures[0] is p1.measures[0] or c1.measures[0] is p2.measures[0]:
print("Child measure has reference to parent measure")
return c1, c2
def one_point(p1, p2):
if constants.NUM_OF_MEASURES == 2:
point1 = 1
else:
point1 = rng.randrange(1, len(p1.measures))
c1 = Individual([], None)
c2 = Individual([], None)
p1_copy = copy.deepcopy(p1)
p2_copy = copy.deepcopy(p2)
for i in range(0, point1):
c1.measures.append(p1_copy.measures[i])
c2.measures.append(p2_copy.measures[i])
for i in range(point1, len(p1_copy.measures)):
c1.measures.append(p2_copy.measures[i])
c2.measures.append(p1_copy.measures[i])
if c1.measures[0] is p1.measures[0] or c1.measures is p2.measures[0]:
print("Child measure has reference to parent measure")
return c1, c2
def create_population(self, n):
'''
Create's an initial population. An individual's genotype is created and retrieved from
GenoTypeFactory. The genotype is initalized to a completely random genome.
'''
population = []
for i in range(n):
genotype = GenotypeFactory.make_genotype(genotype=self.genotype,
config=self.config)
genotype.init_random_genotype(self.genome_length)
individual = Individual(genotype, self.translator)
population.append(individual)
return population
def create(self, genotype=None):
length = self.ann.weights_count() + self.ann.neurons_count() * 3
if genotype is None:
genotype = self.genotype_coder.generate_init_genotype()
phenotype_convertor = BeerTrackerPhenotypeConvertor(
length, self.genotype_coder)
fitness_evaluator = BeerTrackerFitnessEvaluator(self.world, self.ann)
mutation_strategy = BinaryVectorInversionMutation(0.05)
crossover_strategy = FlatlandOnePointCrossover(0.8, length)
return Individual(phenotype_convertor, fitness_evaluator,
mutation_strategy, genotype, crossover_strategy)
def create(self, genotype=None):
length = self.ann.weights_count() + self.ann.biased_neurons_count()
if genotype is None:
# lecture06-slide21: it is recommended to initialize weights from interval [-0,1; 0,1]
vals = [x for x in range(2**self.bits_per_weight) if abs(2*x/(2**self.bits_per_weight)-1) <= 0.1]
genotype = []
for i in range(0, length):
genotype += self.to_binary(random.choice(vals), self.bits_per_weight)
phenotype_convertor = FlatlandPhenotypeConvertor(length)
fitness_evaluator = FlatlandFitnessEvaluator(self.flatland, self.ann)
mutation_strategy = BinaryVectorInversionMutation(0.01)
crossover_strategy = FlatlandOnePointCrossover(0.8, length)
return Individual(phenotype_convertor, fitness_evaluator, mutation_strategy, genotype, crossover_strategy)