Resgression output rather than classification

[atinzad]

Hi,

I was wondering if there is a code modification where the model is used to train for regression rather than classification. I assume that the loss function would be changed, as well as the last layer of the model.

Thank you

[mdeff]

Yep that should be a simple modification. The purpose of this repo is however limited to allow a reproduction of the paper. For new implementations I recommend you to look at PyTorch Geometric and Deep Graph Library.