XOR with Neural Network

XOR: This example is essentially the “Hello World” of neural network programming. This example shows how to construct an neural network to predict the output from the XOR operator. This example uses backpropagation to train the neural network.

a	b	a XOR b
1	1	0
0	1	1
1	0	1
0	0	0

from miml.neural_network import FeedforwardNetwork

indata = array([[0,0],[1,0],[0,1],[1,1]])
ideal = array([0,1,1,0])
net = FeedforwardNetwork(5, 'trainlm')
net.layers[0].set_activation('tanh')
net.train(indata, ideal, isprint=True)
print net.predict([0,0])
print net.predict([0,1])

>>> run script...
Epochs 1: Error=0.666
Epochs 2: Error=0.327
Epochs 3: Error=0.180
Epochs 4: Error=0.018
Epochs 5: Error=0.003
Epochs 6: Error=0.001
array([-0.015162892109416235])
array([0.9941280862563597])