DyGraphs_raw
收藏资源简介:
该数据集是用于论文[CrossLink](https://arxiv.org/pdf/2402.02168.pdf)的动态图数据集。数据集用于学习特定下游图的演化模式,并进行模式特定的链接预测。数据集采用了一种称为条件链接生成(conditioned link generation)的技术,结合了演化和结构建模,以执行演化特定的链接预测。该技术通过transformer解码器架构实现,支持高效的并行训练和推理。数据集包含600万个动态边,涵盖多个领域。实验表明,CrossLink在跨领域链接预测中达到了最先进的性能,平均精度比先进基线提高了11.40%。数据集的格式为CSV,包含源节点、目标节点、交互时间等信息。
This dynamic graph dataset is developed for the paper titled [CrossLink](https://arxiv.org/pdf/2402.02168.pdf). It is designed to learn evolutionary patterns of specific downstream graphs and perform pattern-specific link prediction tasks. The dataset adopts a technique named conditioned link generation, which combines evolutionary and structural modeling to conduct evolution-specific link prediction. This technique is implemented via the Transformer decoder architecture, enabling efficient parallel training and inference. The dataset contains 6 million dynamic edges spanning multiple domains. Experiments demonstrate that CrossLink achieves state-of-the-art performance in cross-domain link prediction, with its average precision outperforming existing state-of-the-art baselines by 11.40%. The dataset is stored in CSV format, including information such as source nodes, target nodes, interaction timestamps and other relevant details.
DyGraphs_raw 数据集概述
基本信息
- 许可证: Apache 2.0
- 标签:
- text
- graph
- 任务类别:
- graph-ml
- 语言:
- en
- 数据格式: csv
数据集简介
该数据集用于论文 CrossLink,旨在学习特定下游图的演化模式,并进行模式特定的链接预测。数据集采用了一种称为条件链接生成的技术,结合了演化和结构建模,以执行演化特定的链接预测。该技术通过transformer解码器架构实现,支持高效的并行训练和推理。数据集包含600万个动态边,涵盖多个领域。在八个未训练的图上进行的广泛实验表明,CrossLink在跨域链接预测方面达到了最先进的性能,平均精度比先进基线提高了11.40%。
数据格式
数据集的格式如下:
| Unnamed: 0 | u | i | ts | label | idx |
|---|---|---|---|---|---|
idx-1 |
source node |
target node |
interaction time |
defalut: 0 |
from 1 to the #edges |
数据可以通过preprocess_data文件夹中的代码进行准备,也可以使用在huggingface上预处理的数据。
引用
如果该工作对您有帮助,请考虑引用:
bibtex @misc{huang2024graphmodelcrossdomaindynamic, title={One Graph Model for Cross-domain Dynamic Link Prediction}, author={Xuanwen Huang and Wei Chow and Yang Wang and Ziwei Chai and Chunping Wang and Lei Chen and Yang Yang}, year={2024}, eprint={2402.02168}, archivePrefix={arXiv}, primaryClass={cs.LG}, url={https://arxiv.org/abs/2402.02168}, }




