https://github.com/graph-com/gssc
[Preprint] Graph State Space Convolution (GSSC)
https://github.com/graph-com/gssc
Last synced: 7 months ago
JSON representation
[Preprint] Graph State Space Convolution (GSSC)
- Host: GitHub
- URL: https://github.com/graph-com/gssc
- Owner: Graph-COM
- Created: 2024-06-07T15:16:43.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2024-06-11T15:24:35.000Z (over 1 year ago)
- Last Synced: 2025-06-23T04:43:59.476Z (8 months ago)
- Language: Python
- Homepage:
- Size: 146 KB
- Stars: 13
- Watchers: 0
- Forks: 2
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
Graph State Space Convolution (GSSC)
This repository contains the official implementation of GSSC as described in the paper: [What Can We Learn from State Space Models for Machine Learning on Graphs?](https://arxiv.org/abs/2406.05815) by Yinan Huang*, Siqi Miao*, and Pan Li.
(*Equal contribution, listed in alphabetical order)
## Installation
All required packages are listed in `environment.yml`.
## Running the code
Replace `--cfg` with the path to the configuration file and `--device` with the GPU device number like below:
```
python main.py --cfg configs/GSSC/peptides-func-GSSC.yaml --device 0 wandb.use False
```
This command will train the model on the `peptides-func` dataset using the GSSC method with default hyperparameters.
## Reproducing the results
We use wandb to log and sweep the results. To reproduce the reported results, one needs to create and login to a wandb account. Then, one can launch the sweep using the configuration files in the `configs` directory.
For example, to reproduce the tuned results of GSSC on the `peptides-func` dataset, one can launch the sweep using `configs/GSSC/peptides-func-GSSC-tune.yaml`.
## Acknowledgement
This repository is built upon [GraphGPS (Rampasek et al., 2022)](https://github.com/rampasek/GraphGPS).