Прогнозирование нейронной сети с нулевой ошибкой

import random
import math
 
 
inputs_number = 4
 
x_array = []
 
for x in range(inputs_number):
    random_value = random.uniform(0, 1)
    x_array.append(random_value)
 
w_array = []
 
for x in range(inputs_number):
    random_value = random.uniform(0, 1)
    w_array.append(random_value)
 
T = random.uniform(0, 1)
 
alpha = random.uniform(0, 1)
 
e = random.uniform(0, 1)
 
list_out = []
list_wanted = []
list_error = []
 
 
def activation(value):
 
    a = 1
    b = 9
    d = 0.5
 
    return a * math.sin(b * value) + d
 
 
def function():
 
    global x_array, w_array, T, alpha, e, list_out, list_wanted, list_error
 
    xw_array = []
 
    for x in range(inputs_number):
        xw_array.append(float(x_array[x] * w_array[x]))
 
    output = 0
 
    for x in range(inputs_number):
        output += xw_array[x]
 
    output -= T
 
    y = float(activation(output))
 
    E = float(0.5 * ((y - e) ** 2))
 
    list_out.append(y)
    list_wanted.append(e)
    list_error.append(E)
 
    for x in range(inputs_number):
        w_array[x] = w_array[x] - alpha * (y - e) * x_array[x]
 
    T = T + alpha * (y - e)
 
    alpha = 1 / (1 + y)
 
 
def forecasting():
 
    global x_array, w_array, T, alpha, e, list_out, list_wanted, list_error
 
    index = 30
    p = 4
 
    xt = 0
 
    for x in range(inputs_number):
        xt += w_array[x] * list_out[index - p + (x + 1) - 1]
 
    list_wanted[index] = list_out[index]
    list_out[index] = xt
    list_error[index] = list_out[index] - list_wanted[index]
    index = index + 1
 
 
for step in range(30):
    function()
 
w_array_new = w_array.copy()
 
for step in range(15):
    function()
 
w_array = w_array_new.copy()
 
for step in range(15):
    forecasting()
 
for step in range(45):
    print(f"{step + 1}\t{list_out[step]}\t{list_wanted[step]}\t{list_error[step]}")