Hierarchical Graph Capsule Network

Authors

  • Jinyu Yang University of Texas at Arlington
  • Peilin Zhao Tencent AI Lab
  • Yu Rong Tencent AI Lab
  • Chaochao Yan University of Texas at Arlington
  • Chunyuan Li University of Texas at Arlington
  • Hehuan Ma University of Texas at Arlington
  • Junzhou Huang University of Texas at Arlington

Keywords:

Graph-based Machine Learning, Bioinformatics, Social Networks, Graph Mining, Social Network Analysis & Community

Abstract

Graph Neural Networks (GNNs) draw their strength from explicitly modeling the topological information of structured data. However, existing GNNs suffer from limited capability in capturing the hierarchical graph representation which plays an important role in graph classification. In this paper, we innovatively propose hierarchical graph capsule network (HGCN) that can jointly learn node embeddings and extract graph hierarchies. Specifically, disentangled graph capsules are established by identifying heterogeneous factors underlying each node, such that their instantiation parameters represent different properties of the same entity. To learn the hierarchical representation, HGCN characterizes the part-whole relationship between lower-level capsules (part) and higher-level capsules (whole) by explicitly considering the structure information among the parts. Experimental studies demonstrate the effectiveness of HGCN and the contribution of each component. Code: https://github.com/uta-smile/HGCN

Downloads

Published

2021-05-18

How to Cite

Yang, J., Zhao, P., Rong, Y., Yan, C., Li, C., Ma, H., & Huang, J. (2021). Hierarchical Graph Capsule Network. Proceedings of the AAAI Conference on Artificial Intelligence, 35(12), 10603-10611. Retrieved from https://ojs.aaai.org/index.php/AAAI/article/view/17268

Issue

Section

AAAI Technical Track on Machine Learning V