Simple Neural Net | Python Fiddle

import numpy as np
iterations = 750
# input patterns. Only 2 inputs 'on' yeilds a True result, everything else would be False
X = np.array([ [0,0,1],[0,1,1],[1,0,1],[1,1,1],[0,0,0],[1,1,0]])
# answers to training examples '1' means True, '0' means false
y = np.array([[0,1,1,0,0,1]]).T

#network structure
# n = number of examples
# i = shape of example, in this case, 3
n,i=X.shape
syn0 = (2*np.random.random((i,n)) - 1)
syn1 = (2*np.random.random((n,1)) - 1)
for j in xrange(iterations):
	# perform the forward pass through the network
	l1 = 1/(1+np.exp(-(np.dot(X,syn0))))
	l2 = 1/(1+np.exp(-(np.dot(l1,syn1))))
	# take the answer given by the net and calculate how wrong its guess is
	l2_delta = (y - l2)*(l2*(1-l2))
	l1_delta = l2_delta.dot(syn1.T) * (l1 * (1-l1))
	# penalize the nuerons that got it 'wrong' and reward the nueron that got it
	# right. in this example we have 3 nuerons that are trying to learn the problem
	# once the network is well trained, the nuerons will fire 'more correctly' and 
	# have less error
	syn1 += l1.T.dot(l2_delta)
	syn0 += X.T.dot(l1_delta)

#uncomment next line to show net examples it has not seen. it will likely get them wrong
# the correct answers would be '0',. You can also make up your own sequence for testing.
# Bonus: modify the traing set above with this two missing examples (be sure to 
# correct answers to Y as well ) and the run again to 'learn' from the two missing values as well

#X=np.array([[1,0,0],[0,1,0]])

# Query the network by showing it the examples. it will output its answer. the raw results
# show how close it is to the actual value. for example, .95 is confidently close to 1
# and .05 is confidently close to 0. This residual can be thought of as the error. the 
# higher the iteration, the lower the error
l1 = 1/(1+np.exp(-(np.dot(X,syn0))))
l2 = 1/(1+np.exp(-(np.dot(l1,syn1))))

print 'raw results:'
print l2.T
print 'Neural net prediction:'
print np.round(l2.T)

Python Fiddle

Python Cloud IDE