PAGTNs #28
Description
This seems like a smart way to handle some of the uneasiness I feel with attentional stuff. This could be in addition to or instead of the GAT proposed in this issue...
Python version, from Regina Barzilay's group:
paper: http://arxiv.org/abs/1905.12712
code: https://github.com/benatorc/PA-Graph-Transformer