-
-
Save RyotaBannai/846667ceddff912cd8c3695c7cc2181e to your computer and use it in GitHub Desktop.
A two layer neural network written in Python, which trains itself to solve a variation of the XOR problem.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from numpy import exp, array, random, dot, reshape | |
from autograd import grad | |
class NeuronLayer(): | |
def __init__(self, neuron_n, inputs_n): | |
self.weights = 2 * random.random((inputs_n, neuron_n)) - 1 | |
class NeuralNetwork(): | |
def __init__(self, layer1, layer2): | |
self.layer1 = layer1 | |
self.layer2 = layer2 | |
sigmoid = lambda x: return 1 / (1 + exp(-x)) | |
sigmoid_deriv = grad(sigmoid) | |
def train(self, t_inputs, t_outputs, iteration_n): | |
for iteration in xrange(iteration_n): | |
# Calculate neural network | |
op_layer1, op_layer2 = self.calculate_output(t_inputs) | |
# Calculate the error for layer 2 | |
# (The difference between the desired output and the predicted output). | |
layer2_error = t_outputs - op_layer2 | |
layer2_delta = layer2_error * self.sigmoid_deriv(op_layer2) | |
# Calculate the error for layer 1 | |
# (By looking at the weights in layer 1, we can determine by how much layer 1 contributed to the error in layer 2). | |
layer1_error = layer2_delta.dot(self.layer2.weights.T) | |
layer1_delta = layer1_error * self.sigmoid_deriv(op_layer1) | |
# Calculate how much to adjust the weights by | |
layer1_adjustment = inputs.T.dot(layer1_delta) | |
layer2_adjustment = op_layer1.T.dot(layer2_delta) | |
# Adjust the weights. | |
self.layer1.weights += layer1_adjustment | |
self.layer2.weights += layer2_adjustment | |
def calculate_output(self, inputs): | |
ot_layer1 = self.sigmoid(dot(inputs, self.layer1.weights)) | |
ot_layer2 = self.sigmoid(dot(output_from_layer1, self.layer2.weights)) | |
return ot_layer1, ot_layer2 | |
# The neural network prints its weights | |
def print_weights(self): | |
print (" Layer 1 (4 neurons, each with 3 inputs): \n") | |
print (self.layer1.weights,'\n') | |
print (" Layer 2 (1 neuron, with 4 inputs): \n") | |
print (self.layer2.weights,'\n') | |
if __name__ == "__main__": | |
random.seed(1) | |
layer1 = NeuronLayer(4, 3) # 4 neurons, each with 3 inputs | |
layer2 = NeuronLayer(1, 4) # a single neuron with 4 inputs | |
neural_network = NeuralNetwork(layer1, layer2) # Combine the layers to create a neural network | |
neural_network.print_weights() # Combine the layers to create a neural network | |
# The training set. We have 7 examples, each consisting of 3 input values | |
# and 1 output value. | |
inputs = reshape([0, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0], (3, -1)) | |
outputs = array([[0, 1, 1, 1, 1, 0, 0]]).T | |
# Train the neural network using the training set. | |
iteration_n = 60000 | |
neural_network.train(inputs, outputs, iteration_n) | |
neural_network.print_weights() | |
# Test the neural network with a new situation. | |
print ("Stage 3) Considering a new situation [1, 1, 0] -> ?: ") | |
hidden_state, output = neural_network.calculate_output(array([1, 1, 0])) | |
print (output) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment