深度学习--梯度下降法

首先是公式的推导，直接看图吧

代码： 

import numpy as np
import matplotlib.pyplot as plt
x_data=[1.0,2.0,3.0]
y_data=[2.0,4.0,6.0]
w=1.0
def forward(x):
    return w*x
def cost(x_data,y_data):
    cost=0
    for x,y in zip(x_data,y_data):
        y_pred=forward(x)
        cost+=(y_pred-y)**2
    return cost/len(x_data)
def gradient(x_data,y_data):
    grad=0
    for x,y in zip(x_data,y_data):
        grad+=2*x*(w*x-y)
    return grad/len(x_data)
print('Predict (before training)',4,forward(4))
a=[]
b=[]
for epoch in range(100):
    cost_val=cost(x_data,y_data)
    grad_val=gradient(x_data,y_data)
    w-=0.01*grad_val
    a.append(epoch)
    b.append(cost_val)
    print('Epoch:',epoch,'w=',w,'loss=',cost_val)
print('Predict (after training)',4,forward(4))
plt.plot(a,b)
plt.show()

随机梯度下降的代码：

#如果画出图像最后损失函数上升，就是训练失败，这时候可以尝试把学习率也就是α调小
#随机梯度下降（可能某一段是平的，你没有办法向前走）
#随机梯度，就是从你的数据中选取一个数据，而不是所有数据的和
import numpy as np
import matplotlib.pyplot as plt
x_data=[1.0,2.0,3.0]
y_data=[2.0,4.0,6.0]
w=1.0
def forward(x):
    return w*x
def cost(x,y):
    y_pred=forward(x)
    return (y_pred-y)**2
def gradient(x,y):
    return 2*(w*x-y)*x
print('Predict (before training)',4,forward(4))
a=[]
b=[]
cnt=1
for epoch in range(100):
    for x,y in zip(x_data,y_data):
        loss=cost(x,y)
        grade=gradient(x,y)
        w=w-0.01*grade
        a.append(cnt)
        cnt+=1
        b.append(loss)
print('Predict (after training)',4,forward(4))
plt.plot(a,b)
plt.show()