C2N-ABDP: Cluster-to-Node Attention-Based Differentiable Pooling

Rongji Ye, Lixin Cui, Luca Rossi, Yue Wang, Zhuo Xu, Lu Bai, Edwin R. Hancock

Research output: Chapter in book / Conference proceedingConference article published in proceeding or bookAcademic researchpeer-review


Graph neural networks have achieved state-of-the-art performance in various graph based tasks, including classification and regression at both node and graph level. In the context of graph classification, graph pooling plays an important role in reducing the number of graph nodes and allowing the graph neural network to learn a hierarchical representation of the input graph. However, most graph pooling methods fail to effectively preserve graph structure information and node feature information when reducing the number of nodes. At the same time, the existing hierarchical differentiable graph pooling methods cannot effectively calculate the importance of nodes and thus cannot effectively aggregate node information. In this paper, we propose an attention-based differentiable pooling method, which aggregates nodes into clusters when reducing the scale of the graph, uses singular value decomposition to calculate cluster information during the aggregation process, and captures node importance information through a novel attention mechanism. The experimental results show that our approach outperforms competitive models on benchmark datasets.

Original languageEnglish
Title of host publicationGraph-Based Representations in Pattern Recognition - 13th IAPR-TC-15 International Workshop, GbRPR 2023, Proceedings
EditorsMario Vento, Pasquale Foggia, Vincenzo Carletti, Donatello Conte
PublisherSpringer Science and Business Media Deutschland GmbH
Number of pages11
ISBN (Print)9783031427947
Publication statusPublished - Sept 2023
Event13th IAPR-TC-15 International Workshop on Graph-Based Representations in Pattern Recognition, GbRPR 2023 - Vietri sul Mare, Italy
Duration: 6 Sept 20238 Sept 2023

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume14121 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference13th IAPR-TC-15 International Workshop on Graph-Based Representations in Pattern Recognition, GbRPR 2023
CityVietri sul Mare


  • Attention
  • Graph neural network
  • Graph pooling

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'C2N-ABDP: Cluster-to-Node Attention-Based Differentiable Pooling'. Together they form a unique fingerprint.

Cite this