TY - GEN
T1 - DeGNN
T2 - 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2021
AU - Miao, Xupeng
AU - Gürel, Nezihe Merve
AU - Zhang, Wentao
AU - Han, Zhichao
AU - Li, Bo
AU - Min, Wei
AU - Rao, Susie Xi
AU - Ren, Hansheng
AU - Shan, Yinan
AU - Shao, Yingxia
AU - Wang, Yujie
AU - Wu, Fan
AU - Xue, Hui
AU - Yang, Yaming
AU - Zhang, Zitao
AU - Zhao, Yang
AU - Zhang, Shuai
AU - Wang, Yujing
AU - Cui, Bin
AU - Zhang, Ce
N1 - Funding Information:
This work is supported by the National Key Research and Development Program of China (No. 2018YFB1004403), the National Natural Science Foundation of China (No. 61832001, U1936104), PKU-Tencent joint research Lab, Beijing Academy of Artificial Intelligence (BAAI), CAAI Huawei MindSpore Open Fund, and The Fundamental Research Funds for the Central Universities 2020RC25. Ce Zhang and the DS3Lab gratefully acknowledge the support from the Swiss National Science Foundation (Project Number 200021_184628), Innosuisse/SNF BRIDGE Discovery (Project Number 40B2-0_187132), European Union Horizon 2020 Research and Innovation Programme (DAPHNE, 957407), Botnar Research Centre for Child Health, Swiss Data Science Center, Alibaba, Cisco, eBay, Google Focused Research Awards, Oracle Labs, Swisscom, Zurich Insurance, Chinese Scholarship Council, and the Department of Computer Science at ETH Zurich.
Publisher Copyright:
© 2021 ACM.
PY - 2021/8/14
Y1 - 2021/8/14
N2 - Mining from graph-structured data is an integral component of graph data management. A recent trending technique, graph convolutional network (GCN), has gained momentum in the graph mining field, and plays an essential part in numerous graph-related tasks. Although the emerging GCN optimization techniques bring improvements to specific scenarios, they perform diversely in different applications and introduce many trial-and-error costs for practitioners. Moreover, existing GCN models often suffer from oversmoothing problem. Besides, the entanglement of various graph patterns could lead to non-robustness and harm the final performance of GCNs. In this work, we propose a simple yet efficient graph decomposition approach to improve the performance of general graph neural networks. We first empirically study existing graph decomposition methods and propose an automatic connectivity-ware graph decomposition algorithm, DeGNN. To provide a theoretical explanation, we then characterize GCN from the information-theoretic perspective and show that under certain conditions, the mutual information between the output after l layers and the input of GCN converges to 0 exponentially with respect to l. On the other hand, we show that graph decomposition can potentially weaken the condition of such convergence rate, alleviating the information loss when GCN becomes deeper. Extensive experiments on various academic benchmarks and real-world production datasets demonstrate that graph decomposition generally boosts the performance of GNN models. Moreover, our proposed solution DeGNN achieves state-of-the-art performances on almost all these tasks.
AB - Mining from graph-structured data is an integral component of graph data management. A recent trending technique, graph convolutional network (GCN), has gained momentum in the graph mining field, and plays an essential part in numerous graph-related tasks. Although the emerging GCN optimization techniques bring improvements to specific scenarios, they perform diversely in different applications and introduce many trial-and-error costs for practitioners. Moreover, existing GCN models often suffer from oversmoothing problem. Besides, the entanglement of various graph patterns could lead to non-robustness and harm the final performance of GCNs. In this work, we propose a simple yet efficient graph decomposition approach to improve the performance of general graph neural networks. We first empirically study existing graph decomposition methods and propose an automatic connectivity-ware graph decomposition algorithm, DeGNN. To provide a theoretical explanation, we then characterize GCN from the information-theoretic perspective and show that under certain conditions, the mutual information between the output after l layers and the input of GCN converges to 0 exponentially with respect to l. On the other hand, we show that graph decomposition can potentially weaken the condition of such convergence rate, alleviating the information loss when GCN becomes deeper. Extensive experiments on various academic benchmarks and real-world production datasets demonstrate that graph decomposition generally boosts the performance of GNN models. Moreover, our proposed solution DeGNN achieves state-of-the-art performances on almost all these tasks.
KW - graph decomposition
KW - graph neural network
KW - information loss
UR - http://www.scopus.com/inward/record.url?scp=85114944257&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85114944257&partnerID=8YFLogxK
U2 - 10.1145/3447548.3467312
DO - 10.1145/3447548.3467312
M3 - Conference contribution
AN - SCOPUS:85114944257
T3 - Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
SP - 1223
EP - 1233
BT - KDD 2021 - Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
PB - Association for Computing Machinery
Y2 - 14 August 2021 through 18 August 2021
ER -