Neural Message Passing for Multi-Relational Ordered and Recursive Hypergraphs

Source code for NeurIPS 2020 paper: Neural Message Passing for Multi-Relational Ordered and Recursive Hypergraphs

Model Usage:

• Please use requirements.txt to install dependencies. To start training, run the following in GMPNN or MPNNR folder.

  python gmpnn.py --data WP-IND --agg max --log False
  python mpnnr.py --data cora --split 1 --log False
  

  • --data denotes the dataset to use
  • --log indicates whether to log results (and dump checkpoints for GMPNN)
  • --agg denotes the type of aggregation for GMPNN (max / mean / sum)
  • --split is the split number to use for MPNNR

Citation:

@incollection{gmpnnr_neurips20,
author = {Naganand Yadati},
title = {Neural Message Passing for Multi-Relational Ordered and Recursive Hypergraphs},
booktitle = {Advances in Neural Information Processing Systems (NeurIPS) 33},
year = {2020},
publisher = {Curran Associates, Inc.}
}