基本元件

1. 卷积

2. batchnorm

loss函数

torch.nn.MSELoss()

>>> a = torch.rand(3) >>> a tensor([0.2161, 0.2227, 0.9175]) >>> b = torch.rand(3) >>> b tensor([0.6976, 0.9149, 0.4918]) >>> mse = torch.nn.MSELOSS() >>> mse(a, b) tensor(0.2974) >>> ((0.2161-0.6976)**2 + (0.2227-0.9149)**2 + (0.9175-0.4918)**2)/3 0.2974011933333333

MSELoss是求高斯距一个函数。

1. 均方误差函数(Mean Squared Eqation)

torch.nn.CrossEntropyLoss()

1.描述两个概率分布间的距离

2.交叉熵函数

import torch.nn as nn loss = nn.CrossEntropyLoss() input = torch.randn(3, 5, requires_grad=True) target = torch.empty(3, dtype=torch.long).random_(5) output = loss(input, target) output.backward()

分类问题用Cross Entropy

回归问题用MSE

基本运算

加减乘除