network,
training_one,
method = "Newton-CG",
ERROR_LIMIT = 1e-4
)
# Train the network using Scaled Conjugate Gradient
scaled_conjugate_gradient(
network,
training_one,
ERROR_LIMIT = 1e-4
)
# Train the network using resilient backpropagation
resilient_backpropagation(
network,
training_one, # specify the training set
ERROR_LIMIT = 1e-3, # define an acceptable error limit
#max_iterations = (), # continues until the error limit is reach if this argument is skipped
# optional parameters
weight_step_max = 50.,
weight_step_min = 0.,
start_step = 0.5,
learn_max = 1.2,
learn_min = 0.5
)
network.print_test( training_one )
training_one, # specify the training set
ERROR_LIMIT=0.001, # define an acceptable error limit
#max_iterations = 100, # continues until the error limit is reach if this argument is skipped
# optional parameters
learning_rate=0.03, # learning rate
momentum_factor=0.4, # momentum
)
# Train the network using SciPy
scipyoptimize(network, training_one, method="Newton-CG", ERROR_LIMIT=1e-4)
# Train the network using Scaled Conjugate Gradient
scaled_conjugate_gradient(network, training_one, ERROR_LIMIT=1e-4)
# Train the network using resilient backpropagation
resilient_backpropagation(
network,
training_one, # specify the training set
ERROR_LIMIT=1e-3, # define an acceptable error limit
#max_iterations = (), # continues until the error limit is reach if this argument is skipped
# optional parameters
weight_step_max=50.,
weight_step_min=0.,
start_step=0.5,
learn_max=1.2,
learn_min=0.5)
network.print_test(lst)
network,
training_one,
method = "Newton-CG",
ERROR_LIMIT = 1e-4
)
# Train the network using Scaled Conjugate Gradient
scaled_conjugate_gradient(
network,
training_one,
ERROR_LIMIT = 1e-4
)
# Train the network using resilient backpropagation
resilient_backpropagation(
network,
training_one, # specify the training set
ERROR_LIMIT = 1e-3, # define an acceptable error limit
#max_iterations = (), # continues until the error limit is reach if this argument is skipped
# optional parameters
weight_step_max = 50.,
weight_step_min = 0.,
start_step = 0.5,
learn_max = 1.2,
learn_min = 0.5
)
network.print_test( lst)
) # Train the network using SciPy #network.scipyoptimize( # training_one, # method = "Newton-CG", # ERROR_LIMIT = 1e-4 # ) ## Train the network using Scaled Conjugate Gradient #network.scg( # training_one, # ERROR_LIMIT = 1e-4 # ) # Train the network using resilient backpropagation #network.resilient_backpropagation( # training_one, # specify the training set # ERROR_LIMIT = 1e-3, # define an acceptable error limit # #max_iterations = (), # continues until the error limit is reach if this argument is skipped # # # optional parameters # weight_step_max = 50., # weight_step_min = 0., # start_step = 0.5, # learn_max = 1.2, # learn_min = 0.5 # ) network.print_test(training_one)
def main(file_path):
input_length = 0
training_one = []
with open(file_path, mode="r") as fh:
temp_data = []
readed_data = csv.reader(fh, delimiter=',')
for row in readed_data:
input_length = len(row) - 2
name = row[0]
for j in range(1, len(row) - 1):
temp_data.append(float(row[j]))
temp_data = normalize(temp_data)
target = bitfield(int(row[len(row) - 1]))
temp_instance = Instance(temp_data, target, name)
temp_data = []
training_one.append(temp_instance)
settings = {
# Required settings
"cost_function":
sum_squared_error,
"n_inputs":
input_length, # Number of network input signals
"layers": [(26, sigmoid_function), (15, sigmoid_function),
(10, sigmoid_function), (3, sigmoid_function)],
# [ (number_of_neurons, activation_function) ]
# The last pair in you list describes the number of output signals
# Optional settings
"weights_low":
-1, # Lower bound on initial weight range
"weights_high":
0, # Upper bound on initial weight range
"save_trained_network":
False, # Whether to write the trained weights to disk
"input_layer_dropout":
0.0, # dropout fraction of the input layer
"hidden_layer_dropout":
0.0, # dropout fraction in all hidden layers
}
# initialize the neural network
network = NeuralNet(settings)
# load a stored network configuration
# network = NeuralNet.load_from_file( "trained_configuration.pkl" )
# Train the network using backpropagation
backpropagation(
network,
training_one, # specify the training set
ERROR_LIMIT=1e-3, # define an acceptable error limit
#max_iterations = 100, # continues until the error limit is reach if this argument is skipped
# optional parameters
learning_rate=0.01, # learning rate
momentum_factor=0.9, # momentum
max_iterations=100000)
"""
# Train the network using SciPy
scipyoptimize(
network,
training_one,
method = "Newton-CG",
ERROR_LIMIT = 1e-4
)
# Train the network using Scaled Conjugate Gradient
scaled_conjugate_gradient(
network,
training_one,
ERROR_LIMIT = 1e-4
)
# Train the network using resilient backpropagation
resilient_backpropagation(
network,
training_one, # specify the training set
ERROR_LIMIT = 1e-3, # define an acceptable error limit
#max_iterations = (), # continues until the error limit is reach if this argument is skipped
# optional parameters
weight_step_max = 50.,
weight_step_min = 0.,
start_step = 0.5,
learn_max = 1.2,
learn_min = 0.5
)
"""
# есть метод save_to_file в neuralnet. надо попробовать его использовать
return network.print_test(training_one)