Нейронная сеть производит классификацию на 4 класса, график(персептрон)

import numpy as np
import matplotlib.pyplot as plt
import random
 
def predict(row,weights,delta):
    activations = [delta[i] + np.dot(weights[i], row[:-1]) for i in range(4)]
    return np.argmax(activations)
 
def train_we(train,lr,n_epoch):
    weights=[np.zeros(len(train[0])-1) for _ in range(4)]
    delta = [0.0]*4
    for epoch in range(n_epoch):
        for row in train:
            prediction = predict(row,weights,delta)
            errors = [0]*4
            errors[row[-1]] = 1
            for i in range(4):
                error = errors[i] - (1 if i == prediction else 0)
                delta[i] = delta[i] + lr * error
                weights[i] += lr * error * np.array(row[:-1])
    return weights, delta
 
 
 
# 5 points for class 0
x1_0 = [0.4, 0.57, 0.62, 0.55, 0.67] 
x2_0 = [0.73, 0.18, 0.62, 0.16, 0.69]
 
# 5 points for class 1
x1_1 = [0.54, 0.32, 0.11, 0.94, 0.33] 
x2_1 = [-0.76,  -0.48, -0.28,  -0.36, -0.33] 
 
# 5 points for class 2
x1_2 = [-0.35, -0.63, -0.95, -0.89, -0.03]
x2_2 = [0.3, 0.15, 0.43, 0.51, 0.18]
 
# 5 points for class 3
x1_3 = [-0.5, -0.45, -0.33, -0.74, -0.47]
x2_3 = [-0.46, -0.04, -0.37,  -0.31,  -0.08]
 
# create dataset
dataset = [[1, x1_0[i], x2_0[i], 0] for i in range(5)] + \
          [[1, x1_1[i], x2_1[i], 1] for i in range(5)] + \
          [[1, x1_2[i], x2_2[i], 2] for i in range(5)] + \
          [[1, x1_3[i], x2_3[i], 3] for i in range(5)]
 
l_rate=0.1 #step of bias
n_epoch=1000 #number of epochs to train
 
weights, delta = train_we(dataset,l_rate,n_epoch)
 
#compare the prediction with what we have chosen
for row in dataset:
    predictit= predict(row,weights,delta)
    print('Выбрано:%d , Предсказано:%d ' % (row[-1],predictit))
 
#dividing points into 4 groups and creating a graph
for row in dataset:
    if row[-1] == 0:
        plt.scatter(row[1], row[2], s=10, c='red')
    elif row[-1] == 1:
        plt.scatter(row[1], row[2], s=10, c='blue')
    elif row[-1] == 2:
        plt.scatter(row[1], row[2], s=10, c='green')
    else:
        plt.scatter(row[1], row[2], s=10, c='yellow')
 
def line(x1,weights, i, j):
    return ((weights[j][1]*x1 - weights[i][1]*x1) / (weights[i][2] - weights[j][2])) - ((weights[i][0] - weights[j][0]) / (weights[i][2] - weights[j][2]));
 
x1_range = np.arange(-3,3,0.5)
plt.plot(x1_range, line(x1_range, weights, 0, 1), color='black')
plt.plot(x1_range, line(x1_range, weights, 2, 3), color='black')
plt.grid(True)
plt.show()