search

hzhao98 / GDCL

Graph Debiased Contrastive Learning with Joint Representation Clustering
21 stars 2 forks source link

求助帖 #1

Closed ZIbuyu1200 closed 1 year ago

ZIbuyu1200 commented 2 years ago

作者大大好， 请问能提供一下你预训练好的文件吗？ 我还想请教一下用MVGRL预训练 谢谢大大Orz

yueliu1999 commented 2 years ago

Hi, I am interested in your paper “Graph Debiased Contrastive Learning with Joint Representation Clustering”. However, I have the same issue with @ZIbuyu1200 when I run your released code. Could you help us.

zzwjames commented 1 year ago

The result of MVGRL cannot be reproduced, how could the author achieves a higher result than that? The code is also hard to run.

hzhao98 commented 1 year ago

作者大大好， 请问能提供一下你预训练好的文件吗？ 我还想请教一下用MVGRL预训练 谢谢大大Orz

Hi, the pretrained models are provided in our code. The pretrained MVGRL model can be reproduced with utilizing the k-means algorithm and the multi-view graph contrastive loss (https://github.com/kavehhassani/mvgrl).

Hi, I am interested in your paper “Graph Debiased Contrastive Learning with Joint Representation Clustering”. However, I have the same issue with @ZIbuyu1200 when I run your released code. Could you help us.

Hi, the pretrained models are provided in our code. The pretrained MVGRL model can be reproduced with utilizing the k-means algorithm and the multi-view graph contrastive loss (https://github.com/kavehhassani/mvgrl).

The result of MVGRL cannot be reproduced, how could the author achieves a higher result than that? The code is also hard to run. The pretrained results of MVGRL can absolutely be reproduced with utilizing the k-means algorithm and the multi-view graph contrastive loss (https://github.com/kavehhassani/mvgrl). And on this basis, utilizing joint clustering and debiased contrastive learning in our method can indeed enhance the performance in our code.