search

google-deepmind / graph_nets

Build Graph Nets in Tensorflow
https://arxiv.org/abs/1806.01261
Apache License 2.0
5.34k stars 783 forks source link

Add optional arguments to graph blocks and modules #122

Closed caiodadauto closed 4 years ago

caiodadauto commented 4 years ago

Dear authors, I added to the library argument options to be used in the evaluation of any module based on the graph blocks. These arguments aim to make possible the use of layers like Dropout or Batch Normalization inside the graph blocks to process either edge, node or global features. I also implemented some tests to validate my implementation, and I change the documentation as needed.

I hope this simple feature could be usefull to improve graph-nets framework.

Best, Caio.

caiodadauto commented 4 years ago

@alvarosg the changes are done.