def __init__(self,
input_count,
output_count,
activation_function=Step(1, 0)):
NeuralNetwork.__init__(self)
self.create_layer(input_count, WeightedSum())
self.create_layer(output_count, WeightedSum(), activation_function)
def __init__(self, neuron_count, y_high, y_low):
"""
:type neuron_count: int
:type y_high: float
:type y_low: float
"""
Layer.__init__(self, neuron_count, WeightedSum(), Step(y_high, y_low))
self.connect_neurons(self)
from synapyse.impl.learning.back_propagation import BackPropagation
from synapyse.impl.multi_layer_perceptron import MultiLayerPerceptron
__author__ = 'Douglas Eric Fonseca Rodrigues'
training_set = TrainingSet(2, 1)
training_set \
.append([0.0, 0.0], [0.0]) \
.append([0.0, 1.0], [1.0]) \
.append([1.0, 0.0], [1.0]) \
.append([1.0, 1.0], [0.0])
n = MultiLayerPerceptron()
n \
.create_layer(2, WeightedSum()) \
.create_layer(3, WeightedSum(), Tanh(2)) \
.create_layer(1, WeightedSum(), Tanh(2))
n.randomize_weights()
b = BackPropagation(n, learning_rate=0.1, max_error=0.01)
b.on_after_iteration = lambda x: print(x.actual_iteration, ':', x.
total_network_error)
b.learn(training_set)
for training_set_row in training_set:
print(n.set_input(training_set_row.input_pattern).compute().output)
# Importa os dados para treinamento
dados_treinamento = TrainingSet(input_count=4, output_count=3) \
.import_from_file('iris_training.data')
# Importa os dados para teste
dados_teste = TrainingSet(input_count=4, output_count=3) \
.import_from_file('iris_testing.data')
# Testa diversas taxas de aprendizagem
for taxa_aprendizado in [0.01, 0.1, 0.3, 0.5, 0.7, 0.9]:
# Cria a rede neural artificial
rede_neural = MultiLayerPerceptron() \
.create_layer(neuron_count=4,
input_function=WeightedSum()) \
.create_layer(neuron_count=3,
input_function=WeightedSum(),
activation_function=Sigmoid()) \
.create_layer(neuron_count=3,
input_function=WeightedSum(),
activation_function=Sigmoid()) \
.randomize_weights()
# Cria o algoritmo de aprendizado
aprendizado = MomentumBackPropagation(neural_network=rede_neural,
learning_rate=taxa_aprendizado,
max_error=0.05,
momentum=0,
max_iterations=30)
from synapyse.impl.learning.back_propagation import BackPropagation
from synapyse.base.learning.training_set import TrainingSet
from synapyse.impl.activation_functions.sigmoid import Sigmoid
from synapyse.impl.input_functions.weighted_sum import WeightedSum
from synapyse.impl.multi_layer_perceptron import MultiLayerPerceptron
__author__ = 'Douglas Eric Fonseca Rodrigues'
# Creating a training_set based in a text file
training_set = TrainingSet(13, 1) \
.import_from_file('heart_disease.txt', ',') \
.normalize()
# Creating and configuring the network
multi_layer_perceptron = MultiLayerPerceptron() \
.create_layer(13, WeightedSum()) \
.create_layer(8, WeightedSum(), Sigmoid()) \
.create_layer(1, WeightedSum(), Sigmoid()) \
.randomize_weights()
# Generating an excel-file result
# Create a workbook and add a worksheet.
workbook = xlsxwriter.Workbook('heart_disease.xlsx')
for learning_rate in [0.01, 0.1, 0.3, 0.5, 0.7]:
print('Running learning_rate =', learning_rate, '...')
# Creating and configuring the learning method
momentum_backpropagation = BackPropagation(
neural_network=multi_layer_perceptron,
.append([1.0, 1.0, 1.0, 0.0, 1.0, 0.0], [1.0, 0.0, 0.0, 0.0, 0.0]) \
.append([1.0, 1.0, 1.0, 1.0, 1.0, 1.0], [1.0, 0.0, 0.0, 0.0, 1.0]) \
.append([1.0, 0.0, 1.0, 1.0, 1.0, 0.0], [1.0, 0.0, 0.0, 1.0, 0.0]) \
.append([0.0, 1.0, 1.0, 0.0, 1.0, 0.0], [1.0, 0.0, 0.0, 1.0, 1.0]) \
.append([0.0, 1.0, 1.0, 1.0, 1.0, 0.0], [1.0, 0.0, 1.0, 0.0, 0.0]) \
.append([1.0, 0.0, 0.0, 0.0, 1.0, 1.0], [1.0, 0.0, 1.0, 0.0, 1.0]) \
.append([1.0, 0.0, 1.0, 0.0, 1.0, 1.0], [1.0, 0.0, 1.0, 1.0, 0.0]) \
.append([0.0, 1.0, 1.0, 1.0, 1.0, 0.0], [1.0, 0.0, 1.0, 1.0, 1.0]) \
.append([1.0, 1.0, 0.0, 0.0, 1.0, 1.0], [1.0, 1.0, 0.0, 0.0, 0.0]) \
.append([1.0, 1.0, 0.0, 1.0, 1.0, 1.0], [1.0, 1.0, 0.0, 0.0, 1.0]) \
.append([1.0, 0.0, 0.0, 1.0, 1.0, 1.0], [1.0, 1.0, 0.0, 1.0, 0.0])
neural_network = MultiLayerPerceptron()
neural_network \
.create_layer(6, WeightedSum()) \
.create_layer(7, WeightedSum(), Sigmoid()) \
.create_layer(5, WeightedSum(), Sigmoid()) \
.randomize_weights()
b = MomentumBackPropagation(neural_network=neural_network,
learning_rate=0.1,
momentum=0.4,
max_error=0.02)
b.on_after_iteration = lambda obj: print(obj.actual_iteration, ':', obj.
total_network_error)
b.learn(training_set)
print('total_error=', b.total_network_error)
from synapyse.impl.multi_layer_perceptron import MultiLayerPerceptron
from synapyse.util.logger import Logger
__author__ = 'Douglas Eric Fonseca Rodrigues'
Logger.enable_logger(Logger.INFO)
sim = 0.3
# Creating a training_set based in a text file
training_set_training, training_set_test = TrainingSet(21, 4) \
.import_from_file('car_evaluation.txt', ',') \
.slice(80)
# Creating and configuring the network
multi_layer_perceptron = MultiLayerPerceptron() \
.create_layer(21, WeightedSum()) \
.create_layer(14, WeightedSum(), Sigmoid()) \
.create_layer(4, WeightedSum(), Sigmoid()) \
.randomize_weights()
# Creating and configuring the learning method
momentum_backpropagation = MomentumBackPropagation(
neural_network=multi_layer_perceptron,
learning_rate=0.3,
momentum=0.6,
max_error=0.001)
# Configuring a log after each learning method iteration
momentum_backpropagation.on_after_iteration = lambda b: print(
b.actual_iteration, ':', b.total_network_error)
from synapyse.impl.activation_functions.linear import Linear
from synapyse.impl.activation_functions.tanh import Tanh
from synapyse.impl.input_functions.weighted_sum import WeightedSum
from synapyse.impl.learning.back_propagation import BackPropagation
from synapyse.impl.multi_layer_perceptron import MultiLayerPerceptron
__author__ = 'Douglas Eric Fonseca Rodrigues'
training_set = TrainingSet(2, 1)
training_set.append([0.0, 1.0], [1.0])
n = MultiLayerPerceptron()
n \
.create_layer(2, WeightedSum(), Linear()) \
.create_layer(2, WeightedSum(), Tanh(2)) \
.create_layer(2, WeightedSum(), Tanh(2)) \
.create_layer(1, WeightedSum(), Tanh(2))
n00 = n.layers[0].neurons[0]
n01 = n.layers[0].neurons[1]
b0 = n.layers[0].neurons[2]
n10 = n.layers[1].neurons[0]
n11 = n.layers[1].neurons[1]
b1 = n.layers[1].neurons[2]
n20 = n.layers[2].neurons[0]
n21 = n.layers[2].neurons[1]
b2 = n.layers[2].neurons[2]