About
Recently, the advances in deep graph learning have enabled a large amount of applications related to graph-structured data, from recommendation system, social network analysis to novel molecule design. Deep graph generation, which brings unprecedented opportunities in generating/modeling/designing new graph structures (e.g. molecules) and understanding the learnt representation by deep graph learning models. However, it faces huge challenges, such as generalizability to different types of graphs, modeling graphs with constraints, etc. We observe one of the most important challenges which prevent the advances of the community is the lack of data in multiple dimensions, which includes but not limited to, of various types, of various domains, of various application areas. To reach this goal, we release GraphGT, a systematic graph generation and transformation dataset collection. We collect, process, and open-source diverse types of graph-structured data to help the community advance graph generative models for various types of graphs.Datasets
- Molecular Graphs [Chemistry] [Detail] [QM9] [ZINC] [MOSES] [ChEMBL] [Molecule_Optimization] [Chemical_Reaction]
- Proteins [Biology] [Detail] [Protein]
- Brain Networks [Biology] [Detail] [Brain]
- Physical Simulation Networks [Physics] [Detail] [Physical_Simulation]
- Collaboration Networks [Social Science] [Detail] [Collaboration]
- Traffic Networks [Operational Research and Management Science] [Detail] [TraffNet-LA] [TraffNet-BAY]
- Authentication Networks [Electrical Engineering] [Detail] [Authentication]
- IoT Networks [Electrical Engineering] [Detail] [IoT]
- Skeleton Graphs [Artificial Intelligence] [Detail] [Kinetics] [NTU]
- Synthetic Networks [Detail] [Scale-Free] [Erdos-Renyi] [Barabasi-Albert] [Waxman] [Random-Geometric]
- Circuit Networks [Electrical Engineering] [Detail] [Link]
- Social Networks [Social Science] [Detail] [Link]
- Semantic Graphs [Artificial Intelligence] [Detail] [Link]
- Scene Graphs [Artificial Intelligence] [Detail] [Link]
Tasks
Citation
Please cite us if you use GraphGT in your work:
@inproceedings{du2021graphgt,
title={GraphGT: Machine Learning Datasets for Graph Generation and Transformation},
author={Du, Yuanqi and Wang, Shiyu and Guo, Xiaojie and Cao, Hengning and Hu, Shujie and Jiang, Junji and Varala, Aishwarya and Angirekula, Abhinav and Zhao, Liang},
booktitle={NeurIPS 2021},
year={2021}
}
Yuanqi Du, Shiyu Wang, Xiaojie Guo, Hengning Cao, Shujie Hu, Junji Jiang, Aishwarya Varala, Abhinav Angirekula, and Liang Zhao. GraphGT: Machine learning datasets for deep graph generation and transformation. In Thirty-fifth Conference on Neural Information Processing Systems, 2021.