Wuyxin / DIR-GNN

Official code of "Discovering Invariant Rationales for Graph Neural Networks" (ICLR 2022)

Home Page:https://arxiv.org/abs/2201.12872

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

parameters in paper

syn13355061266 opened this issue · comments

Hi, I would like to ask you what are the parameters of the spurious-motif dataset in your paper, i.e. the parameters in the graph_stats(base_num) function under the path spmotif_gen/spmotif.ipynb.

Hi the parameters we used should be the default values in the notebook file

Thank you for your reply. I noticed that your notebook code only generates a training set of 3,000 graphs, so you just need to change it to generate a training set of 9,000 graphs under the original parameters and leave the other parameters untouched, right?

Yes exactly!