Learning with batches of graphs

[zjsteeve]

Hello @tkipf,

Thank you so much for making this work available to the public. It has been incredibly helpful in my research.

Currently in train.py, it appears that a single graph instance is being passed to the session in each epoch. I am wondering if there is a way to learn on batches of graphs, in the same way that you can pass batches of images to a traditional CNN? Thank you for any help you may be able to offer!

[tkipf]

Batch-wise training should be possible using the approach described in this thread: https://github.com/tkipf/gcn/issues/4 -- you might have to explicitly exclude edges between different graphs in the decoder depending on what setting you have in mind.