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

实现和论文对不上 #4

Open
ownnaruto opened this issue Sep 9, 2024 · 1 comment
Open

实现和论文对不上 #4

ownnaruto opened this issue Sep 9, 2024 · 1 comment

Comments

@ownnaruto
Copy link

ownnaruto commented Sep 9, 2024

你好,你这个源码实现和论文中的描述好像不太一致

论文中提到:
"The source feature network is pre-trained with masked autoencoding [14] to reconstruct masked patches using unmasked ones in a long input sequence ... "
然而实际上代码中source feature network并没有maksed autoencoding部分

并且训练node feature network时,论文中给出的loss如下:
截屏2024-09-09 10 07 19

代码的实现中还加入了Ldistil(x_vt)

@KL4805
Copy link
Owner

KL4805 commented Sep 9, 2024

你好,代码里没有给出source feature network的pre-train代码,因为我们直接使用了STEP (https://github.com/GestaltCogTeam/STEP) 的代码。
这部分代码对应于训练target node feature network。如果target存在一定的长期时间序列的话,我们可以加入Ldistill(x_vt),代码中是这么实现的。论文里没有写是因为我们不能假设所有目标域都有长期时间序列。

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