#!/usr/bin/python
from pyfann import fann
connection_rate = 1
learning_rate = 0.7
num_input = 2
num_neurons_hidden = 4
num_output = 1
desired_error = 0.0001
max_iterations = 100000
iterations_between_reports = 1000
ann = fann.create(connection_rate, (num_input, num_neurons_hidden, num_output))
ann.set_learning_rate(learning_rate)
ann.set_activation_function_output(fann.SIGMOID_SYMMETRIC_STEPWISE)
ann.train_on_file("../../examples/xor.data", max_iterations, iterations_between_reports, desired_error)
ann.save("nets/xor_float.net")
def print_callback(epochs, error):
print "Epochs %8d. Current MSE-Error: %.10f\n" % (epochs, error)
return 0
# initialize network parameters
connection_rate = 1
learning_rate = 0.7
num_neurons_hidden = 32
desired_error = 0.000001
max_iterations = 300
iterations_between_reports = 1
# create training data, and ann object
print "Creating network."
train_data = fann.read_train_from_file(os.path.join("..","..","benchmarks","datasets","mushroom.train"))
ann = fann.create(connection_rate, (train_data.get_num_input(), num_neurons_hidden, train_data.get_num_output()))
ann.set_learning_rate(learning_rate)
# start training the network
print "Training network"
ann.set_activation_function_hidden(fann.SIGMOID_SYMMETRIC_STEPWISE)
ann.set_activation_function_output(fann.SIGMOID_STEPWISE)
ann.set_training_algorithm(fann.TRAIN_INCREMENTAL)
ann.train_on_data(train_data, max_iterations, iterations_between_reports, desired_error)
# test outcome
print "Testing network"
test_data = fann.read_train_from_file(os.path.join("..","..","benchmarks","datasets","mushroom.test"))
