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

为什么去掉Conv-TasNet的skip-connection部分? #41

Open
StarrySilence1 opened this issue Mar 3, 2022 · 2 comments
Open

为什么去掉Conv-TasNet的skip-connection部分? #41

StarrySilence1 opened this issue Mar 3, 2022 · 2 comments

Comments

@StarrySilence1
Copy link

想知道Conv-TasNet为什么不保留原论文中的skip-connection部分呢?

@StarrySilence1
Copy link
Author

最近在尝试改动Conv-TasNet,看了github上不同作者的实现,发现都没有使用论文中的skip-connnection。自己在加入skip-connection之后发现网络就很难收敛了。
另一方面:最近在研究Conv-TasNet是否能通过不基于mask,而是直接生成语音的方式来分离音频。基于cln和gln会损失输入的尺度信息,我把所有LN层都改为了BN层,同时依照提出BN层的论文,BN层应该放在activation函数之前,将Normalization层提前了。同时也保留了skip-connection。发现网络基本不收敛,想请教一下是否有什么明显的问题?

@JusperLee
Copy link
Owner

  1. 加入skip-connection:性能应该不会下降太多。
    2.对于regression的方法,我没有做过这种尝试,你可以尝试将encoder后的feature与估计的feature相加试一下。

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

2 participants