How to get and use the node features of the syn-cora dataset?

[LirongWu]

How to get and use the node features of the syn-cora dataset?

[LirongWu]

Thank you very much for your patient answer, which solved my problem.

[jiong-zhu]

Hi Lirong, thank you for your interest in our work!

• Yes, the downloaded dataset should include everything you need to run the experiments, including node features. The .allx files we provided should be in the same format as the ones provided by Kipf et al.
• On the level of train-validation-test splits (e.g. workspace/bfd0ae58a3c870b860ab42382ac416c2/features/313255294c48f167af526a058ba65d62/splits/33cffc3d560873eb6d85e2640d956fd4/), the files are already organized using mostly the same format as Planetoid and Kipf et al.
• However, if you want to directly use the dataset code provided by Kipf et al., you should include the corresponding change as in the Line 188 of this file on their code, otherwise you may not get correct results on the synthetic graphs without correct sorting of the node ID. A more recommended way is to use our dataset code, which is compatible to the one provided by Kipf et al. (See Line 67-68 here for an example)