About Accuracy rate

[songshuhan]

Hello author, I want to do some comparative tests with the model provided by Deeprobust, but I find that the results are quite different from those in the paper. For example, when I attack 0.25 data set with meta, the results of Prognn are only 51%, and the results of GCN are only 43%. GCNJaccard is only 55%, basically the other data sets are also very different, I don't know if it's me or the parameter setting

[songshuhan]

On cora dataset

[ChandlerBang]

Hi, have you tried the script in https://github.com/ChandlerBang/Pro-GNN/blob/master/scripts/meta/cora_meta.sh?

[songshuhan]

Thanks for your reply. I have tried a lot of seeds and models that can be imported from Deeprobust. Now I can get the corresponding results of some of the models in the paper. However, when GCNJaccard meta attack 25% on pubmed, it will never run 83%, only 75%. I don't know how to use GCNJaccard for comparison HAHAHAHAHA , can you give me a seed that runs to 83%?

[ChandlerBang]

Hi, you may try tune the threshold in GCNJaccard in the range of {0.01, 0.02, 0.03, 0.04, 0.05, 0.1}. This hyper-parameter is important to its performance.

[zhongjian-zhang]

Hello author, I want to do some comparative tests with the model provided by Deeprobust, but I find that the results are quite different from those in the paper. For example, when I attack 0.25 data set with meta, the results of Prognn are only 51%, and the results of GCN are only 43%. GCNJaccard is only 55%, basically the other data sets are also very different, I don't know if it's me or the parameter setting

hello, I meet the same question with you, could you tell me how to address it? thank you