DyGSSM

DyGSSM: Multi-view Dynamic Graph Embeddings with State Space Model Gradient Update

This repository is our PyTorch implementation of DyGSSM.

How to run

Create and activate environment

conda create -p dygssm_env python=3.8.10 -y
conda activate ./dygssm_env
pip install torch==2.2.0 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
pip install torch-scatter -f https://data.pyg.org/whl/torch-2.2.0+cu118.html
pip install torch-sparse -f https://data.pyg.org/whl/torch-2.2.0+cu118.html
pip install torch-cluster -f https://data.pyg.org/whl/torch-2.2.0+cu118.html
pip install dgl-cu117 -f https://data.dgl.ai/wheels/repo.html

Install the python dependencies

pip install -r requirements.txt

Clone HawkesGNN, ROLAND, and WinGNN code

1 - Download roland code

git clone https://github.com/snap-stanford/roland.git
git clone https://github.com/oncemoe/hawkesGNN.git

2 - Move files from extra_ to their coresponding folders

cd extra
mv * ../roland/run
cd ../

3 - Add ROLAND

cd roland
pip install -e .
get_roland_public_data.sh will generate the public dataset folder and download the datasets
cd ../

4 - run

python main.py --dataset uci-msg --epoch 1 --repeat 1

Acknowledgement