Minimal neural network

minimal-nn.py

import numpy as np

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

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

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

np.random.seed(1)

# hidden layer filled with random
h = 2*np.random.random((3,1))-1

# training
for i in range(10000):
    
    # forward propаgation
    l0 = X
    l1 = nonlin(np.dot(l0, h))

    # error calc
    l1_err = y - l1
    l1_dlt = l1_err * nonlin(l1, True)
    
    # correcting hidden layer 
    h += np.dot(l0.T, l1_dlt)


print("result of training\n", l1)
print("must be\n", y)

test = np.array([[0,0,0],[1,1,0]])
print("test\n", nonlin(np.dot(test,h)))