刘二大人 《PyTorch深度学习实践》第3讲 梯度下降算法

指路 ☞《PyTorch深度学习实践》完结合集_哔哩哔哩_bilibili

知识补充：

1、预设一个w值，然后算斜率（梯度），如果梯度大于0，我们要往梯度小的方向进行，即减去它，反之一样。α是学习率，此处设为0.01

 2、梯度下降法只能算出局部最优，没办法得到全局最优

3、鞍点，梯度为0

 

4、如果cost随epoch曲线上升了，原因可能是α取大了；加权均值可以让曲线更圆滑

 

 5、随机梯度下降公式

 6、梯度下降与随机梯度下降之间的区别

**************************************************************************************************************

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 x * w
def cost(xs, ys):
    cost = 0
    for x, y in zip(xs, ys):
        y_p = forward(x)
        cost += (y_p - y)**2
        return cost / len(xs)
def gradient(xs, ys):
    grad = 0
    for x, y in zip(xs, ys):
        grad = 2 * x * (x * w - y)
        grad += grad
    return grad / len(xs)
print('before training', 4, forward(4))

epoch_list = []
cost_list = []
for epoch in range(100):
    cost_val = cost(x_data, y_data)
    grad_val = gradient(x_data, y_data)
    w -= 0.01 * grad_val
    print('Epoch=', epoch, 'w={:.2f}'.format(w), 'cost={:.2f}'.format(cost_val))
    epoch_list.append(epoch)
    cost_list.append(cost_val)
print('after training', 4, '{:.2f}'.format(forward(4)))

plt.plot(epoch_list, cost_list)
plt.xlabel('Epoch')
plt.ylabel('Cost')
plt.show()

部分运行结果如下：

 

*************************************************************************************************************

 更改为随机梯度下降，即将cost 改为loss

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 x*w
def loss(x, y):
    y_p = forward(x)
    return (y_p - y)**2
def gradient(x, y):
    return 2 * x * (x*w - y)


print('before training', 4, forward(4))

epoch_list = []
loss_list = []
for epoch in range(100):
    for x, y in zip(x_data, y_data):
        grad = gradient(x, y)
        w -= 0.01 * grad
        print('\tgrad:', x, y, grad)
        l = loss(x, y)

    print('Epoch=', epoch, 'w={:.2f}'.format(w), 'loss={:.2f}'.format(l))
    epoch_list.append(epoch)
    loss_list.append(l)
print('after training', 4, '{:.2f}'.format(forward(4)))

plt.plot(epoch_list, loss_list)
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.show()

部分运行结果：