Skip to content

Commit

Permalink
Update README.md
Browse files Browse the repository at this point in the history
  • Loading branch information
Hanson0910 authored Oct 29, 2021
1 parent 350bb5f commit 8eed2a3
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -57,7 +57,7 @@
1. [交叉熵Loss,多分类任务中通常用到交叉熵Loss作为代价函数,分类交叉熵损失求导更简单,损失仅与正确类别的概率有关](https://github.com/Hanson0910/DL-Algorithm-Summary/blob/main/loss%E7%AF%87/%E5%88%86%E7%B1%BBloss.md)
2. [Focal-Loss,Focal-Loss能够更加关注难例](https://github.com/Hanson0910/DL-Algorithm-Summary/blob/main/loss%E7%AF%87/%E5%88%86%E7%B1%BBloss.md)
3. [Gradient Harmonized Single-stage Detector,分析了单阶段目标检测中的样本不均衡问题,通过梯度分布来动态为不同样本分配不同的梯度。这样原本分类中占据大量梯度的负例和easy-sample就会减少它们的梯度,占据少量的very-hard sample虽然样本比较少,但是却单个梯度很大,这样也会减小它们的梯度,medium-sample就会相对增加梯度](https://github.com/Hanson0910/DL-Algorithm-Summary/blob/main/%E7%9B%AE%E6%A0%87%E6%A3%80%E6%B5%8B%E7%AF%87/Anchor-Base/one-stage/GHM.md)
4. [Bi-Tempered Logistic Loss, 可以在一定程度上解决错误样本标记带来的问题,通过改进log函数和exp函数,添加两个tempered变量减轻错误样本的影响](https://github.com/Hanson0910/DL-Algorithm-Summary/blob/main/%E7%9B%AE%E6%A0%87%E6%A3%80%E6%B5%8B%E7%AF%87/Anchor-Base/one-stage/GHM.md)
4. [Bi-Tempered Logistic Loss, 可以在一定程度上解决错误样本标记带来的问题,通过改进log函数和exp函数,添加两个tempered变量减轻错误样本的影响](https://github.com/Hanson0910/DL-Algorithm-Summary/blob/main/loss%E7%AF%87/%E5%88%86%E7%B1%BBloss.md)

## 回归Loss
1. [L1 Loss, 不论对于什么样的输入值,都有着稳定的梯度,不会导致梯度爆炸问题,对离群点不敏感,具有较为稳健性的解](https://github.com/Hanson0910/DL-Algorithm-Summary/blob/main/loss%E7%AF%87/%E5%9B%9E%E5%BD%92%E6%8D%9F%E5%A4%B1%E5%87%BD%E6%95%B0.md)
Expand Down

0 comments on commit 8eed2a3

Please sign in to comment.