theano-xor.py

from __future__ import print_function

import theano
import theano.tensor as T
import numpy as np
import time

X = theano.shared(value=np.asarray([[0, 1], [1, 0], [0, 0], [1, 1]]), name='X')
y = theano.shared(value=np.asarray([[0], [0], [1], [1]]), name='y')
rng = np.random.RandomState(1234)
LEARNING_RATE = 0.1

def layer(*shape):
    mag = 4. * np.sqrt(6. / sum(shape))
    return theano.shared(value=np.asarray(rng.uniform(low=-mag, high=mag,
    	   size=shape), dtype=theano.config.floatX), name='W', borrow=True, strict=False)

W1 = layer(2, 5)
W2 = layer(5, 1)
b1 = layer(5)
b2 = layer(1)

output = T.nnet.sigmoid(T.dot(T.nnet.relu(T.dot(X, W1) + b1), W2) + b2)
cost = T.mean((y - output) ** 2)
updates = [(W1, W1 - LEARNING_RATE * T.grad(cost, W1)),
           (W2, W2 - LEARNING_RATE * T.grad(cost, W2)),
           (b1, b1 - LEARNING_RATE * T.grad(cost, b1)),
           (b2, b2 - LEARNING_RATE * T.grad(cost, b2))]

train = theano.function(inputs=[], outputs=[], updates=updates)
test = theano.function(inputs=[], outputs=cost)

print('Error before:', test())
start = time.time()
for i in range(10000):
    train()
end = time.time()
print('Error after:', test())
print('Time (s):', end - start)

Error before: 0.372680536564
Error after: 0.000388681638475
Time (s): 0.184696912766