ga = Genetic_Algorithm(individuals = 1000,
mutation_rate = 0.1,
# (4 caracteristicas + 1 do bias) * 2 neuronios = 10
# (2 saidas de neuronio + 1 do bias) * 3 neuronios de saida = 9
# total de pesos 10+9 = 19
genotype = 19,
fenotype = fenotype,
genotype_type = Genetic_Algorithm.TYPE_FLOAT)
# ga.print_individuals()
ga.execute(generations = 100, log=True)
# ga.print_individuals()
print "Best Individual for all generations", ga.get_best_individual().get_genotype(), ga.get_best_individual_fenotype()
genotype = ga.get_best_individual().get_genotype()
pos = 0
for i, layer in enumerate(mlp.get_layers()):
for j, neuron in enumerate(layer.get_neurons()):
end_neuron_position = pos + len(neuron.get_weights())
neuron.get_weights()[:] = genotype[pos: end_neuron_position]
neuron.set_bias(genotype[end_neuron_position])
pos = end_neuron_position + 1
mlp.get_weights('pesos_ga_iris.mlp')
xor = numpy.ndarray((4, 1))
xor[:, 0] = numpy.logical_xor(inputs[:, 0], inputs[:, 1])
return inputs, xor
def test(mlp):
inputs = numpy.ndarray((4, 2))
inputs[:, 0] = numpy.array([0, 0, 1, 1])
inputs[:, 1] = numpy.array([0, 1, 0, 1])
xor = numpy.ndarray((4, 1))
xor[:, 0] = numpy.logical_xor(inputs[:, 0], inputs[:, 1])
for inp, x in zip(inputs, xor):
print inp, x, mlp.output(inp)
if __name__ == '__main__':
inputs, xor = createData()
mlp = MLP(n=0.3, features=2, topology=[4, 2, 1])
errors = mlp.train_data(train=inputs, target=xor, epochs=10000, log=True)
print errors
# print mlp.get_weights()
test(mlp)
mlp.get_weights('xor.mlp')
mutation_rate=0.1,
# (4 caracteristicas + 1 do bias) * 2 neuronios = 10
# (2 saidas de neuronio + 1 do bias) * 3 neuronios de saida = 9
# total de pesos 10+9 = 19
genotype=19,
fenotype=fenotype,
genotype_type=Genetic_Algorithm.TYPE_FLOAT)
# ga.print_individuals()
ga.execute(generations=100, log=True)
# ga.print_individuals()
print "Best Individual for all generations", ga.get_best_individual(
).get_genotype(), ga.get_best_individual_fenotype()
genotype = ga.get_best_individual().get_genotype()
pos = 0
for i, layer in enumerate(mlp.get_layers()):
for j, neuron in enumerate(layer.get_neurons()):
end_neuron_position = pos + len(neuron.get_weights())
neuron.get_weights()[:] = genotype[pos:end_neuron_position]
neuron.set_bias(genotype[end_neuron_position])
pos = end_neuron_position + 1
mlp.get_weights('pesos_ga_iris.mlp')
return inputs, xor
def test(mlp):
inputs = numpy.ndarray((4,2))
inputs[:,0] = numpy.array([0,0,1,1])
inputs[:,1] = numpy.array([0,1,0,1])
xor = numpy.ndarray((4,1))
xor[:,0] = numpy.logical_xor(inputs[:,0], inputs[:,1])
for inp, x in zip(inputs, xor):
print inp, x, mlp.output(inp)
if __name__ == '__main__':
inputs, xor = createData()
mlp = MLP(n = 0.3,
features = 2,
topology = [4,2,1])
errors = mlp.train_data(train = inputs, target = xor, epochs = 10000, log=True)
print errors
# print mlp.get_weights()
test(mlp)
mlp.get_weights('xor.mlp')