Neural network

nn.py

import numpy

# Initialize not-so-randomness
numpy.random.seed(1)

# Sigmoid function : R -> [0-1]
def sigmoid(x, deriv=False):
	return x*(1-x) if deriv else 1/(1+numpy.exp(-x))

# Inputs
X = numpy.array([
	[0,0,1],
	[0,1,1],
	[1,0,1],
	[1,1,1]
])
# Outputs
Y = numpy.array([[0,1,1,0]]).T

# Synapses
syn0 = 2*numpy.random.random((3,4)) - 1
syn1 = 2*numpy.random.random((4,1)) - 1

for j in xrange(60000):
	l0 = X
	l1 = sigmoid(numpy.dot(l0, syn0))
	l2 = sigmoid(numpy.dot(l1, syn1))

	# l2 correction
	l2_error = Y - l2
	l2_delta = l2_error * sigmoid(l2, True)
	syn1 += l1.T.dot(l2_delta)

	# l1 correction
	l1_error = l2_delta.dot(syn1.T)
	l1_delta = l1_error * sigmoid(l1, True)
	syn0 += l0.T.dot(l1_delta)

	if j % 10000 == 0:
		print 'Error : %s' % numpy.mean(numpy.abs(l2_error))
print l2, syn0, syn1

Taken from http://iamtrask.github.io/2015/07/12/basic-python-network/?