validation_set = []
validation_target = []
for i in range(100):
x = -5.0+i*0.1
z = m.exp(-0.5*x*x)/m.sqrt(2.0*m.pi)
validation_set.append([x])
validation_target.append(z)
validation_set = np.array(validation_set)
validation_target = np.array(validation_target)
"""
Build Neural network
"""
net = NeuralNet()
net.build_network(1,4,1, hidden_type="tanh", out_type="linear", scope="regression", verbose=True)
"""
Another way of building the same network
"""
#net = NeuralNet()
#net.add_layer("input",1)
#net.add_layer("sigmoid",4)
#net.add_layer("linear",1)
#net.set_scope("regression")
#net.print_network_structure()
"""
Training
"""
# set parameters for training
validation_labels = []
dataset = list(read(dataset="testing", path=""))
#print len(dataset)
for i in range(len(dataset)):
label, vector = dataset[i]
validation_labels.append(label)
validation_set.append(vector.flatten())
validation_set = np.array(validation_set)
validation_labels = np.array(validation_labels)
"""
Build Neural network
"""
net = NeuralNet()
net.build_network(784,
100,
10,
hidden_type="tanh",
out_type="softmax",
scope="classification")
"""
Training
"""
# set parameters for training
net.set_training_param(learning_rate=0.01,
momentum=0.9,
return_error=True,
batchsize=10,
training_rounds=1)
# train
train_error = net.trainOnDataset(training_set, training_labels)
"""
Print error during training on file
for j in range(100):
y = -5.0 + j * 0.1
vector = [x, y]
z = m.exp(-0.5 * (x * x + y * y)) / 2.0 * m.pi
validation_set.append(vector)
validation_target.append(z)
validation_set = np.array(validation_set)
validation_target = np.array(validation_target)
"""
Build Neural network
"""
net = NeuralNet()
net.build_network(2,
8,
8,
1,
hidden_type="sigmoid",
out_type="linear",
scope="regression",
verbose=True)
"""
Another way of building the same network
"""
#net = NeuralNet()
#net.add_layer("input",2)
#net.add_layer("sigmoid",8)
#net.add_layer("sigmoid",8)
#net.add_layer("linear",1)
#net.set_scope("regression")
#net.print_network_structure()
"""
Training
for i in range(1000):
training_set = training_set + validation_set
training_label = training_label + validation_label
validation_set = np.array(validation_set)
validation_label = np.array(validation_label)
training_set = np.array(training_set)
training_label = np.array(training_label)
"""
Build Neural network
"""
net = NeuralNet()
net.build_network(2,
4,
1,
hidden_type="tanh",
out_type="sigmoid",
scope="classification",
verbose=True)
"""
Another way of building the same network
"""
#net = NeuralNet()
#net.add_layer("input",2)
#net.add_layer("tanh",4)
#net.add_layer("sigmoid",1)
#net.set_scope("classification")
#net.print_network_structure()
"""
Training
"""
