This is my experimental version of the Graph Attention Network (GAT) with two improvement.

1. I try to improve GAT with higher order interaction between nodes of the edge instead of the first order approximation used in the origin version.
2. A minibatch training procedure is also adopted to scale GAT for large dataset.
   The model was first presented by Veličković et. al (2017, https://arxiv.org/abs/1710.10903).

The repo has been forked initially from https://github.com/Diego999/pyGAT.git.