Numpy realizes the classification of iris by perceptron

import random

import matplotlib.pyplot as plt
import numpy as np
from sklearn import datasets
def sign(x):
    if x >= 0:
        return 1
    else:
        return -1
def get_loss(dataset,target,w):
    results = [np.matmul(w.T,data) for data in dataset]
    ''' Predictive value '''
    predict = [sign(i) for i in results]
    ''' Misclassification point '''
    error_index = [i for i in range(len(target)) if target[i] != predict[i]]
    print(' And then there were {} Error points '.format(len(error_index)))
    ''' Calculation loss'''
    loss = 0
    for i in error_index:
        loss += results[i] * (predict[i] - target[i])
    return loss
def get_w(dataset,target,w,learning_rate):
    results = [np.matmul(w.T,data) for data in dataset]
    ''' Predictive value '''
    predict = [sign(i) for i in results]
    ''' Misclassification point '''
    error_index = [i for i in range(len(target)) if target[i] != predict[i]]
    ''' Select random misclassification points '''
    index = random.choice(error_index)
    ''' Stochastic gradient descent '''
    w = w - learning_rate * (predict[index] - target[index]) * dataset[index]
    return w
if __name__ == '__main__':
    ''' Calyx length 、 Calyx width 、 Petal length 、 Petal width '''
    '''0,1 Divisible , Select two categories '''
    dataset, target = datasets.load_iris()['data'], datasets.load_iris()['target']
    target = target[:100]
    target = np.array([1 if i == 1 else -1 for i in target])
    dataset = dataset[:100]
    dataset = np.array([np.array(data[:2]) for data in dataset])
    dataset = np.hstack((np.ones((dataset.shape[0], 1)), dataset))
    w = np.random.rand(len(dataset[0]))
    loss = get_loss(dataset,target,w)
    print(loss)
    epoch = 10000
    learning_rate = 0.1
    for i in range(epoch):
        ''' to update w'''
        w = get_w(dataset,target,w,learning_rate)
        ''' Calculation loss'''
        loss = get_loss(dataset,target,w)
        print(' after {} iteration ,loss：{}'.format(i,loss))
        if loss == 0:
            break
    print(w)
    x_point = np.array([i[1] for i in dataset])
    y_point = np.array([i[2] for i in dataset])
    x = np.linspace(4,7.2,10)
    y = (-w[1] * x - w[0])/w[2]
    plt.scatter(x_point[:50],y_point[:50],label = '0')
    plt.scatter(x_point[-50:],y_point[-50:],label = '1')
    plt.legend()  #  Show Legend 
    plt.plot(x,y)
    plt.show()

Here is a random gradient descent

jie The result of perceptron classification is not unique