简单用python实现几种梯度下降

from matplotlib import pyplot as plt
import  random

 
#生成数据
def data():
    x = range(10)
    y = [(2*i+4) for i in x]
    for i in range(10):
        y[i] = y[i]+random.randint(0,8)-4
    return x,y
 
 
#使用随机梯度下降训练
def SGD(x,y):
    error0 = 0
    step_size = 0.001
    esp = 1e-6
    #a = random.randint(0,4)
    #b = random.randint(0,8)
    a = 1.2  #将给a，b随机赋初始值
    b = 3.5
    m = len(x)
    n = 0
    while True:
        i = random.randint(0,m-1)
        print(i)
        sum0 = a * x[i] + b - y[i]
        sum1 = (a * x[i] + b - y[i])*x[i]
        error1 = (a * x[i] + b - y[i])**2  #计算模型和结果的误差
 
        a = a - sum1*step_size/m
        b = b - sum0*step_size/m
        print('a=%f,b=%f,error=%f'%(a,b,error1))
 
        if abs(error1-error0)<esp:  #误差很小，可以终止迭代
            break
        error0 = error1
        n = n+1
        if n%20==0:
            print('第%d次迭代'%n)
        if (n>500):
            break
    return a,b
if __name__ == '__main__':
    x,y = data()
    a,b = SGD(x,y)
    X = range(10)
    Y = [(a*i+b) for i in X]
 
    plt.scatter(x,y,color='red')
    plt.plot(X,Y)
    plt.show()

	

	
	
	
	
import  matplotlib.pyplot as plt
import random
import matplotlib
 
#生成数据
def data():
    x = range(10)
    y = [(2*s+4) for s in x]
    for i in range(10):
        y[i] = y[i]+random.randint(0,8)-4
    return x, y
 
#使用梯度下降进行训练
def diedai(x,y):
    flag = True
    a = random.randint(0,5)
    b = random.randint(0,10)
    m = len(x)
    arf = 0.005 #学习率
    n = 0
    sum1 = 0
    sum2 = 0
    exp = 0.000001
    error0 = 0
    error1 = 0
    while flag:
 
        for i in range(m):  #计算对应的偏导数
            sum1 = a*x[i]+b-y[i]
            sum2 = (a*x[i]+b-y[i])*x[i]
            error1 = (a*x[i]+b-y[i])**2
        a = a - sum2*arf/m  #对a，b进行更新
        b = b - sum1*arf/m
 
        if