Official Implementation of CopulaLSP

This is an official implementation repository for CopulaLSP. Please refer to our accepted paper for details:
A Scalable Inter-edge Correlation Modeling in CopulaGNN for Link Sign Prediction, ICLR 2026. [paper]

Setup

Install

pip install -r requirements.txt

Details

# python>=3.10
torch==2.7.0
torch-geometric==2.6.1
scikit-learn==1.6.1
hydra-core==1.3.2
hydra-optuna-sweeper==1.2.0
pytorch-memlab
tabulate

Run

BitcoinAlpha

python main.py -m --config-name CopulaLSP \
    seed='range(2017, 2027)' \
    dataset=BitcoinAlpha \
    model.eps=0.04 \
    model.eta=0.0008

BitcoinOTC

python main.py -m --config-name CopulaLSP \
    seed='range(2017, 2027)' \
    dataset=BitcoinOTC \
    model.eps=0.05 \
    model.eta=0.0001

WikiElec

python main.py -m --config-name CopulaLSP \
    seed='range(2017, 2027)' \
    dataset=WikiElec \
    model.eps=0.04 \
    model.eta=0.005

WikiRfa

python main.py -m --config-name CopulaLSP \
    seed='range(2017, 2027)' \
    dataset=WikiRfa \
    model.eps=0.04 \
    model.eta=0.01

SlashDot

python main.py -m --config-name CopulaLSP \
    seed='range(2017, 2027)' \
    dataset=SlashDot \
    model.eps=0.04 \
    model.eta=0.01

Epinions

python main.py -m --config-name CopulaLSP \
    seed='range(2017, 2027)' \
    dataset=Epinions \
    model.eps=0.04 \
    model.eta=0.001

BibTex

@inproceedings{
    CopulaLSP,
    title={A Scalable Inter-edge Correlation Modeling in {C}opula{GNN} for Link Sign Prediction},
    author={Jinkyu Sung and Myunggeum Jee, and Joonseok Lee},
    booktitle={In Proceeding of The International Conference on Learning Representations (ICLR)},
    year={2026}
}