https://github.com/vijaydwivedi75/gnn-lspe

Source code for GNN-LSPE (Graph Neural Networks with Learnable Structural and Positional Representations), ICLR 2022
https://github.com/vijaydwivedi75/gnn-lspe

attention geometric-deep-learning gnn gnn-lspe graph-deep-learning graph-neural-networks graph-representation-learning graph-transformer graphs lspe message-passing molecules positional-encoding representation-learning transformer-networks transformers

Last synced: 3 days ago
JSON representation

Source code for GNN-LSPE (Graph Neural Networks with Learnable Structural and Positional Representations), ICLR 2022

• Host: GitHub
• URL: https://github.com/vijaydwivedi75/gnn-lspe
• Owner: vijaydwivedi75
• License: mit
• Created: 2021-09-13T06:00:28.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2022-02-10T08:18:34.000Z (almost 3 years ago)
• Last Synced: 2024-12-13T03:34:26.029Z (10 days ago)
• Topics: attention, geometric-deep-learning, gnn, gnn-lspe, graph-deep-learning, graph-neural-networks, graph-representation-learning, graph-transformer, graphs, lspe, message-passing, molecules, positional-encoding, representation-learning, transformer-networks, transformers
• Language: Python
• Homepage: http://arxiv.org/abs/2110.07875
• Size: 267 KB
• Stars: 247
• Watchers: 4
• Forks: 35
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        


# Graph Neural Networks with 
 Learnable Structural and Positional Representations




Source code for the paper "**[Graph Neural Networks with Learnable Structural and Positional Representations](https://openreview.net/pdf?id=wTTjnvGphYj)**" by Vijay Prakash Dwivedi, Anh Tuan Luu, Thomas Laurent, Yoshua Bengio and Xavier Bresson, at the **Tenth International Conference on Learning Representations (ICLR) 2022**.

We propose a novel GNN architecture in which the structural and positional representations are decoupled, and are learnt separately to learn these two essential properties. The architecture, named **MPGNNs-LSPE** (MPGNNs with **L**earnable **S**tructural and **P**ositional **E**ncodings), is generic that it can be applied to any GNN model of interest which fits into the popular 'message-passing framework', including Transformers.

![MPGNNs-LSPE](./docs/gnn-lspe.png)  




## 1. Repo installation

[Follow these instructions](./docs/01_repo_installation.md) to install the repo and setup the environment.




## 2. Download datasets

[Proceed as follows](./docs/02_download_datasets.md) to download the benchmark datasets.




## 3. Reproducibility 

[Use this page](./docs/03_run_codes.md) to run the codes and reproduce the published results.




## 4. Reference 

:page_with_curl: Paper [on arXiv](https://arxiv.org/abs/2110.07875)  

:movie_camera: Video by [@vijaydwivedi75](https://github.com/vijaydwivedi75) [on YouTube](https://youtu.be/fft2Q0jEWi0)   

:movie_camera: Video by [@xbresson](https://github.com/xbresson) [on YouTube](https://youtu.be/hADjUl4ymoQ)   

```

@inproceedings{dwivedi2022graph,

  title={Graph Neural Networks with Learnable Structural and Positional Representations},

  author={Vijay Prakash Dwivedi and Anh Tuan Luu and Thomas Laurent and Yoshua Bengio and Xavier Bresson},

  booktitle={International Conference on Learning Representations},

  year={2022},

  url={https://openreview.net/forum?id=wTTjnvGphYj}

}

```