search

wyx11112 / GCMAE

The code for "Generative and Contrastive Paradigms Are Complementary for Graph Self-Supervised Learning"
5 stars 0 forks source link

About reproduce the reported results. #1

Open AliothF opened 7 months ago

AliothF commented 7 months ago

Hello, thank you very much for making your source code available, I'm very interested in your work, the experimental results in the article are impressive, when I ran as requested with --use_cfg in the command to reproduce the reported results, The results on most data sets are very different from the results in the article. Could you further disclose your experimental environment, or are there any hyperparameters that have not been adjusted well?

wyx11112 commented 7 months ago

Thank you for your attention. We have revised some hyperparameters, we will update the related files soon.

------------------ Original ------------------ From: AliothF @.> Date: Tue,Feb 27,2024 11:03 AM To: wyx11112/GCMAE @.> Cc: Subscribed @.***> Subject: Re: [wyx11112/GCMAE] About the reproducible results (Issue #1)

Hello, thank you very much for making your source code available, I'm very interested in your work, the experimental results in the article are impressive, when I ran as requested with --use_cfg in the command to reproduce the reported results, The results on most data sets are very different from the results in the article. Could you further disclose your experimental environment, or are there any hyperparameters that have not been adjusted well?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>

AliothF commented 7 months ago

Thank you very much for your prompt reply, I'm looking forward to the results of running the updated hyperparameters.

AliothF commented 6 months ago

不好意思，再次打扰，请问关于图分类的数据集的超参数大概什么时候能更新？

Charliechen1 commented 2 months ago

Hello Alioth, just to cross validate, I got the following result: with the given hyper parameters, we got 84.45±0.34, 72.56±0.70, 77.27±0.77 separately on Cora, CiteSeer and PubMed dataset with random seed from 0 to 19 (mean and std for 20 random seeds).