Directed or Undirected? #69
-
|
Hi PyGOD I am having trouble figuring out if the models are undirected or directed. I am currently using GCNAE, AnomalyDAE, DOMINANT and AdONE with MLPAE as a baseline. If I understand correctly, the GCN layer in the ''semi-supervised classification with GCNs" paper by Kipf and Welling is currently not supporting directed graphs and this is the paper you are basing your GCN layers upon. However, for some of the other papers they support directed graphs and the dataset you are using, 'WEIBO' is directed. I have a naturally directed graph - is it transformed into an undirected one by your models or is it still directed? Have a nice holiday
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
Hi Natasha. Thank you for using the models in our library as the baselines. Our model is based on PyG, which supports both directed and undirected graphs. As for the GCN layers, they are able to work with directed graphs as well (see this PyG issue). It really depends on the scenario, whether to transform it for better results. You can try both and adopt the better one. |
Beta Was this translation helpful? Give feedback.
Hi Natasha. Thank you for using the models in our library as the baselines. Our model is based on PyG, which supports both directed and undirected graphs. As for the GCN layers, they are able to work with directed graphs as well (see this PyG issue). It really depends on the scenario, whether to transform it for better results. You can try both and adopt the better one.