import numpy as np
import matplotlib.pyplot as plt
import random
#数据个数
M=10
#用于学习的数据x
x=np.array(range(10))
y0=np.random.rand(10)*5
y1=np.array(range(10,59,5))
#用于学习的数据y
y=y0+y1
#学习率
LEARNING_RATE = 0.01
w = 0
b = 0
#假定拟合函数:y=b+wx
def hypothesis(x):
global b,w
return b + w * x
#学习过程
def learn(x,y,alpha):
global b,w
bSum=0
wSum=0
#分别计算b和w的梯度
for i in range(M):
bSum+=(hypothesis(x[i])-y[i])
wSum+=(hypothesis(x[i])-y[i])*x[i]
#利用梯度下降寻找到最拟合的b和w值
b=b-(alpha/M)*bSum
w=w-(alpha/M)*wSum
#学习迭代次数
MAX_ITER=10000
for i in range(MAX_ITER):
learn(x,y,LEARNING_RATE)
print(i,"b:",b,"w:",w)
#绘图可视化学习的效果
y_learned=b+w*x
plt.scatter(x,y,color="red")
plt.plot(x,y_learned)
plt.show()
此例中数据拟合的效果如图: