This repository contains the official implementation of the 'Non-Backtracking Graph Neural Network.' Our code evaluates the performance of NBA-GNNs on various benchmarks, including graph classification, graph regression, and node classification tasks. We have organized the code into two parts: one for the Long-Range Graph Benchmark (LRGB) and the other for citation networks and heterophilic datasets. Please be aware that different environment requirements apply to each dataset.
This codebase is written for python3. To install requirements, you need to choose one of the folders and run:
conda env create --file env.yaml
To train the model please run:
python main.py
This project is licensed under the terms of the MIT license.