In [1]:

    

import numpy as np


    

In [2]:

    

def nonlin(x,deriv=False):
    if(deriv==True):
        return x*(1-x)
    return 1/(1+np.exp(-x))


    

In [3]:

    

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


    

In [4]:

    

X


    

    
Out[4]:





array([[0, 1, 1],
       [1, 1, 1],
       [1, 0, 0]])

In [5]:

    

y = np.array([[0,1,1]]).T


    

In [6]:

    

y


    

    
Out[6]:





array([[0],
       [1],
       [1]])

In [7]:

    

np.random.seed(1)


    

In [8]:

    

syn0 = 2*np.random.random((3,1)) - 1


    

In [14]:

    

syn0


    

    
Out[14]:





array([[-0.16595599],
       [ 0.44064899],
       [-0.99977125]])

In [15]:

    

for iter in xrange(10000):

    # forward prop
    l0 = X
    l1 = nonlin(np.dot(l0,syn0))

    # error
    l1_error = y - l1

    # back prop
    l1_delta = l1_error * nonlin(l1,True)

    # weights update
    syn0 += np.dot(l0.T,l1_delta)


    

In [16]:

    

l0


    

    
Out[16]:





array([[0, 1, 1],
       [1, 1, 1],
       [1, 0, 0]])

In [17]:

    

l1


    

    
Out[17]:





array([[ 0.01142146],
       [ 0.98979175],
       [ 0.99988086]])

In [18]:

    

l1_error


    

    
Out[18]:





array([[-0.01142146],
       [ 0.01020825],
       [ 0.00011914]])

In [19]:

    

l1_delta


    

    
Out[19]:





array([[ -1.28959798e-04],
       [  1.03144553e-04],
       [  1.41932018e-08]])

In [20]:

    

syn0


    

    
Out[20]:





array([[ 9.0351758 ],
       [-1.51020277],
       [-2.95062301]])

In [21]:

    

print "Trained output:"
print l1


    

    




Trained output:
[[ 0.01142146]
 [ 0.98979175]
 [ 0.99988086]]

In [ ]: