87 lines
1.9 KiB
Python
Executable File
87 lines
1.9 KiB
Python
Executable File
#!/bin/python3
|
|
import numpy as np
|
|
import random
|
|
|
|
cicli = 59990
|
|
hidden_layer = 1
|
|
epsilon = 0.01
|
|
|
|
def fdt(x,deriv=False):
|
|
if(deriv==True):
|
|
return (1-np.tanh(x)**2)
|
|
|
|
return np.tanh(x)
|
|
|
|
def error_func(output):
|
|
e=0
|
|
for i in range(len(output)):
|
|
e += (output[i]-Y[i])**2
|
|
return 0.5*e
|
|
|
|
#input data
|
|
X = np.array([[0,1], # primo input
|
|
[0,1],
|
|
[1,1],
|
|
[1,1]])
|
|
|
|
#output data
|
|
Y = np.array([[0],
|
|
[0],
|
|
[1],
|
|
[1]])
|
|
|
|
np.random.seed(1) #per avere sempre lo stesso
|
|
|
|
#sinapsi tra input e Perceptron di output
|
|
syn0 = 0.1*(2*np.random.random((2,hidden_layer)) -1)
|
|
|
|
#print(syn0) #2x1
|
|
|
|
# fase forward
|
|
l0 = X
|
|
l1 = fdt(np.dot(l0,syn0)) #output layer hidden, 4x4 4 esempi, 4 neuroni di output
|
|
|
|
for i in range(cicli):
|
|
print(str(int(100*i/cicli))+"---------------------------------------") #% sul numero totale di cicli
|
|
# fase forward
|
|
|
|
old_error = error_func(l1)
|
|
|
|
#fase backward
|
|
#calcolo delta_nu
|
|
delta_nu = Y - l1
|
|
|
|
#print(syn0[:,0])
|
|
sum = 0
|
|
for nu in range(len(X)): #ciclo tra gli esempi
|
|
sum += epsilon*(delta_nu[nu]*fdt(np.dot(X[nu],syn0),True))*X[nu]
|
|
#print(sum)
|
|
syn0[:,0] += sum
|
|
|
|
l1 = fdt(np.dot(l0,syn0)) #output layer hidden, 4x4 4 esempi, 4 neuroni di output
|
|
|
|
current_error = error_func(l1)
|
|
print(old_error - current_error)
|
|
|
|
if (np.linalg.norm(sum) < 10**-8):
|
|
print("Very low variation in weight-space - exiting")
|
|
print("delta_w :"+str(np.linalg.norm(sum)))
|
|
break
|
|
if (error_func(l1) < 0.001):
|
|
print("Very low error - exiting")
|
|
break
|
|
if (abs(current_error - old_error) < 10**-8):
|
|
print("Very low error variation - exiting")
|
|
break
|
|
|
|
|
|
print("Error: "+str(error_func(l1)))
|
|
for i in range(len(np.transpose(l1))):
|
|
print(np.transpose(l1)[i])
|
|
#for i in range(cicli):
|
|
#l1 = fdt(np.dot(l0, syn0))
|
|
#print(l1)
|
|
#l2 = fdt(np.dot(l1, syn1))
|
|
#print(l2)
|
|
|