当前位置：网站首页>Video 36 Chapter 13

Video 36 Chapter 13

2022-07-22 00:03:00 【I want to send SCI】

 

 

 

 

 

 

 

 

 

 

 

  There are fewer parameters The degree of responsibility has become lower The model is small It can be considered as a stronger regularization effect

RGB Channel mean and variance do narmalize

pytorch Medium normalize application _ Yongli's blog -CSDN Blog _normalize pytorchtorch.nn.functional.normalize(input, p=2, dim=1, eps=1e-12, out=None) among ,p Representation norm ( Here is 2 norm ),dim Represents the dimension of the calculation norm ( The default is 1),eps It's to prevent the denominator from being 0;pytorch Medium normalize In essence, a function is to normalize a dimension , Formula for ：ν=νmax(∣∣ν∣∣p,ϵ)\nu = \frac{\nu}{max(||\nu||_p,\epsilon)}ν=max(∣∣ν∣∣p​,ϵ)ν​ Example a = F.softmax(https://blog.csdn.net/sinat_38685124/article/details/121317832

 

 

  The last level is ten times the learning rate

That number is the learning rate  

原网站

版权声明
本文为[I want to send SCI]所创，转载请带上原文链接，感谢
https://yzsam.com/2022/202/202207210018194843.html

边栏推荐

猜你喜欢

随机推荐