【PyTorch深度学习】第三讲：反向传播

最新推荐文章于 2025-12-16 16:32:45 发布

原创最新推荐文章于 2025-12-16 16:32:45 发布 · 139 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#深度学习 #机器学习 #python

PyTorch 深度学习专栏收录该内容

11 篇文章

订阅专栏

该博客通过一个简单的线性方程 y = w * x 展示了如何在 PyTorch 中使用自动求导机制（backward 函数）进行梯度计算，并应用梯度下降法更新权重 w。代码示例中，初始化权重 w，设置计算梯度，定义损失函数并迭代训练，最终实现了权重的优化，使得预测值逐渐接近真实值。

部署运行你感兴趣的模型镜像

1.数学含义

y = w * x

2.含义解释

使用pytorch的内置backward函数进行自动求导

3.代码演示

import torch

x_data = [1.0,2.0,3.0]
y_data = [2.0,4.0,6.0]


w = torch.tensor([1.0])  #  版本选择
w.requires_grad = True  #是否计算梯度

def forward(x):
    return x * w

def loss(x,y):
    y_pred = forward(x)
    return (y_pred-y) ** 2



print('Predict (before train)', 4,forward(4).item())

for epoch in range(100):
    l = loss(1, 2)#初始化参数，避免出错
    for x,y in zip(x_data,y_data):
        l = loss(x,y)
        l.backward()
        print('\t',x,y,w.grad.item())
        w.data = w.data -0.01 * w.grad.data
        w.grad.data.zero_()

    print("progress:",epoch,l.item())


print('Predict (before train)', 4,forward(4).item())```

4.结果展示

Predict (before train) 4 4.0
1.0 2.0 -2.0
2.0 4.0 -7.840000152587891
3.0 6.0 -16.228801727294922
progress: 0 7.315943717956543
1.0 2.0 -1.478623867034912
2.0 4.0 -5.796205520629883
3.0 6.0 -11.998146057128906
progress: 1 3.9987640380859375
1.0 2.0 -1.0931644439697266
2.0 4.0 -4.285204887390137
3.0 6.0 -8.870372772216797
···············
···············
···············
progress: 95 9.094947017729282e-13
1.0 2.0 -7.152557373046875e-07
2.0 4.0 -2.86102294921875e-06
3.0 6.0 -5.7220458984375e-06
progress: 96 9.094947017729282e-13
1.0 2.0 -7.152557373046875e-07
2.0 4.0 -2.86102294921875e-06
3.0 6.0 -5.7220458984375e-06
progress: 97 9.094947017729282e-13
1.0 2.0 -7.152557373046875e-07
2.0 4.0 -2.86102294921875e-06
3.0 6.0 -5.7220458984375e-06
progress: 98 9.094947017729282e-13
1.0 2.0 -7.152557373046875e-07
2.0 4.0 -2.86102294921875e-06
3.0 6.0 -5.7220458984375e-06
progress: 99 9.094947017729282e-13
Predict (before train) 4 7.999998569488525