Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

what is the meaning of 'f' and 'ff' ? 0.5??? #7

Open
henbucuoshanghai opened this issue Feb 19, 2019 · 7 comments
Open

what is the meaning of 'f' and 'ff' ? 0.5??? #7

henbucuoshanghai opened this issue Feb 19, 2019 · 7 comments

Comments

@henbucuoshanghai
Copy link

what is the meaning of 'f' and 'ff' ? 0.5???

@qiyang77
Copy link

Hello @henbucuoshanghai , If you know the answer, please tell me

@sycophant-stone
Copy link

total loss = (λ/(1+λ))softmax loss +(1/(1+λ)) a-softmax loss
i) 左边是原始softmax loss. 在训练初期这个λ值很大, 因此原始softmax Loss占主要贡献.
ii) λ是和iter有关的(递减), 因此后期λ值会变小(clip到5). 因此后期网络的L-Softmax Loss会占主导.

@henbucuoshanghai
Copy link
Author

厉害。、、为什么需要原始的softmax

@sycophant-stone
Copy link

  1. 可以去看下large softmax loss 论文.Large-Margin Softmax Loss for Convolutional Neural Networks

  2. 简单讲, L-softmax 不好收敛. 开头先用softmax loss. 然后随着epoch增加,逐步增大L-softmax loss的作用.

  3. 引用:

For optimization, normally the stochastic gradient descent will work well. However, when training data has too many subjects (such as CASIA-WebFace
dataset), the convergence of L-Softmax will be more difficult than softmax loss.

@henbucuoshanghai
Copy link
Author

henbucuoshanghai commented Apr 29, 2019 via email

@LuisKay
Copy link

LuisKay commented May 27, 2019

@sycophant-stone 我注意到在Loss_ASoftmax.py中,最后返回的logits是原始logits,而不是updated_logits。实际上如果返回updated_logits和其他几位实现的一样的话,在计算每一个batch的分类正确率时很不正常...那到底应该返回哪一个呢?

@jiazhen-code
Copy link

@sycophant-stone 我注意到在Loss_ASoftmax.py中,最后返回的logits是原始logits,而不是updated_logits。实际上如果返回updated_logits和其他几位实现的一样的话,在计算每一个batch的分类正确率时很不正常...那到底应该返回哪一个呢?

我在另一个代码实现中看到他返回的是updated_logits,但是updated_logits必须喂入标签数据,我如果想单纯进行预测,不知道标签数据该怎么办。我试了试将另一篇的代码改成返回原始logits,结果准确率远远低于返回updated_logits

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants