1. 連續(xù)值損失函數(shù) 

總結(jié)：主要使用胡貝兒損失函數(shù)，應(yīng)用于連續(xù)數(shù)值的預(yù)測之間的誤差損失，參考地址

import torch
import torch.nn as nna = torch.tensor([[1, 2], [3, 4]], dtype=torch.float)
b = torch.tensor([[3, 5], [8, 6]], dtype=torch.float)loss_fn1 = torch.nn.MSELoss(reduction='none')
loss1 = loss_fn1(a, b)           # loss1 是按照原始維度輸出，即對應(yīng)位置的元素相減然后求平方                
print(loss1)  # 輸出結(jié)果：tensor([[4., 9.], [25., 4.]])loss_fn2 = torch.nn.MSELoss(reduction='sum')
loss2 = loss_fn2(a, b)          # loss2 是所有位置的損失求和
print(loss2)  # 輸出結(jié)果：tensor(42.)loss_fn3 = torch.nn.MSELoss(reduction='mean')
loss3 = loss_fn3(a, b)          # loss3 是所有位置的損失求和后取平均
print(loss3)  # 輸出結(jié)果：tensor(10.5000)

2. 二分類損失函數(shù)