0
点赞
收藏
分享

微信扫一扫

Flutter第七弹 网格列表GridView

1.参数

loss_func_none = nn.CrossEntropyLoss(reduction="none")
loss_func_mean = nn.CrossEntropyLoss(reduction="mean")
loss_func_sum = nn.CrossEntropyLoss(reduction="sum")

默认是“mean”  也就是说当loss_func_none = nn.CrossEntropyLoss()时 会输出一组batch 的损失平均值

import torch
import torch.nn as nn
loss_func = nn.CrossEntropyLoss(reduction="none")

pre = torch.tensor([[0.8, 0.5, 0.2, 0.5],
                         [0.2, 0.9, 0.3, 0.2],
                         [0.4, 0.3, 0.7, 0.1],
                         [0.1, 0.2, 0.4, 0.8]], dtype=torch.float)
tgt_index = torch.tensor([0,1,2,3], dtype=torch.long)
print(loss_func(pre, tgt_index))

输出如下

import torch
import torch.nn as nn
loss_func = nn.CrossEntropyLoss()

pre = torch.tensor([[0.8, 0.5, 0.2, 0.5],
                         [0.2, 0.9, 0.3, 0.2],
                         [0.4, 0.3, 0.7, 0.1],
                         [0.1, 0.2, 0.4, 0.8]], dtype=torch.float)
tgt_index = torch.tensor([0,1,2,3], dtype=torch.long)
print(loss_func(pre, tgt_index))

输出

tgt表示样本类别的真实值,有两种表示形式,一种是类别的index,另一种是one-hot形式。

tgt_index_data = torch.tensor([0,
                               1,
                               2,
                               3], dtype=torch.long)
tgt_onehot_data = torch.tensor([[1, 0, 0, 0],
                                [0, 1, 0, 0],
                                [0, 0, 1, 0],
                                [0, 0, 0, 1]], dtype=torch.float)

 损失函数|交叉熵损失函数 (zhihu.com)

 

 2.计算过程

loss_func = nn.CrossEntropyLoss()
pre = torch.tensor([0.8, 0.5, 0.2, 0.5], dtype=torch.float)
tgt = torch.tensor([1, 0, 0, 0], dtype=torch.float)
print("手动计算:")
print("1.softmax")
print(torch.softmax(pre, dim=-1))
print("2.取对数")
print(torch.log(torch.softmax(pre, dim=-1)))
print("3.与真实值相乘")
print(-torch.sum(torch.mul(torch.log(torch.softmax(pre, dim=-1)), tgt), dim=-1))
print()
print("调用损失函数:")
print(loss_func(pre, tgt))

 

 交叉熵损失函数会自动对输入模型的预测值进行softmax。因此在多分类问题中,如果使用nn.CrossEntropyLoss(),则预测模型的输出层无需添加softmax层。

参考torch.nn.CrossEntropyLoss() 参数、计算过程以及及输入Tensor形状 - 知乎 (zhihu.com)

举报

相关推荐

0 条评论