search

Shen-Lab / GraphCL

[NeurIPS 2020] "Graph Contrastive Learning with Augmentations" by Yuning You, Tianlong Chen, Yongduo Sui, Ting Chen, Zhangyang Wang, Yang Shen
MIT License
548 stars 103 forks source link

Question about Unsupervised-TU ''test'' and ''val'' #65

Open jiaruHithub opened 1 year ago

jiaruHithub commented 1 year ago

Hi! In Unsupervised-TU, such as logistic regression classifier, there are two parts, test and val, annotated in the ten fold cross verification section. But in my opinion, these two parts are trained separately on two fully connected layer models using the same training set and test set. Am I understanding this correctly? What is the purpose of this? Does it have any special meaning?

Thanks!

yyou1996 commented 1 year ago

Hi @jiaruHithub,

I do not fully understand your question. Would you be able to elaborate more on it? The setting of unsupervised_tu follows https://github.com/sunfanyunn/InfoGraph/tree/master/unsupervised without changes.