CrossLink

🚀 Introduction

CrossLink learns the evolution pattern of a specific downstream graph and subsequently makes pattern-specific link predictions. It employs a technique called conditioned link generation, which integrates both evolution and structure modeling to perform evolution-specific link prediction. This conditioned link generation is carried out by a transformer-decoder architecture, enabling efficient parallel training and inference. CrossLink is trained on extensive dynamic graphs across diverse domains, encompassing 6 million dynamic edges. Extensive experiments on eight untrained graphs demonstrate that CrossLink achieves state-of-the-art performance in cross-domain link prediction. Compared to advanced baselines under the same settings, CrossLink shows an average improvement of 11.40% in Average Precision across eight graphs. Impressively, it surpasses the fully supervised performance of 8 advanced baselines on 6 untrained graphs.

🛠️ Prerequisites

Environment

conda create -n your_env_name python=3.8
conda activate crosslink
pip install -r requirements.txt

Dataset

Please keep the dataset in the fellow format:

Unnamed: 0	u	i	ts	label	idx
idx-1	source node	target node	interaction time	defalut: 0	from 1 to the #edges

You can prepare those data by the code in preprocess_data folder

You can also use our precessed data in huggingface

💡 Usage

Train with 6 garph

python train.py

Eval:

python eval.py

our code is build based on DyGLib and keep the same settings.