SGAT

SGAT (Simplicial Graph Attention Network) is graph neural model for heterogeneous graph datasets. This repo supplements our paper published in IJCAI-22. This version of code is specifically for IMDB dataset.

Setup

Experiments tested on python3.9 with cuda 10.2 and dgl-cuda 0.6.0.

How to run

Parameters

There are two parameters to explicitly change in train_sgat.py file.

• UNINFORMATIVE : Set to True to run with Random Node Features (RNF)
• EDGE_FEATURES : Set to True to run SGAT-EF else it will be SGAT.

First, create conda virtual environment with the following command

conda create --name <env> --file requirements.txt

Running the code for IMDB dataset (with GPU)

python train_sgat.py --dataset IMDB --L 10 --lr 0.005 --num_heads 2 --hidden_units 64 --weight_decay 0.0005