1. import numpy as np
2. X = np.array(([2,9], [1,5], [3,6]), dtype = float)
3. y = np.array(([.92], [.86], [.89]), dtype = float)
4. X = X/np.amax(X, axis = 0)
5.  
6. def sigmoid(x):
7.     return 1/(1+np.exp(-x))
8.  
9. def der_sigmoid(x):
10.     return x*(1-x)
11.  
12. epoch = 5000
13. lr = 0.01
14. neurons_i = 2
15. neurons_h = 3
16. neurons_o = 1
17.  
18. weight_h = np.random.uniform(size=(neurons_i, neurons_h))
19. bias_h = np.random.uniform(size=(1, neurons_h))
20. weight_o = np.random.uniform(size=(neurons_h, neurons_o))
21. bias_o = np.random.uniform(size=(1, neurons_o))
22.  
23. for i in range(epoch):
24.     inp_h = np.dot(X, weight_h) + bias_h
25.     out_h = sigmoid(inp_h)
26.    
27.     inp_o = np.dot(out_h, weight_o) + bias_o
28.     out_o = sigmoid(inp_o)
29.    
30.     err_o = y - out_o
31.     grad_o = der_sigmoid(out_o)
32.     delta_o = err_o * grad_o
33.    
34.     err_h = delta_o.dot(weight_o.T)
35.     grad_h = der_sigmoid(out_h)
36.     delta_h = err_h * grad_h
37.    
38.     weight_o += out_h.T.dot(delta_o) * lr
39.     weight_h += X.T.dot(delta_h) * lr
40.    
41. print('Input:\n', X)
42. print('Actual:\n', y)
43. print('Predicted:\n', out_o)