Rethinking pooling in graph neural networks
Diego Mesquita, Amauri H. Souza Jr., Samuel Kaski
Abstract
Graph pooling is a central component of a myriad of graph neural network (GNN) architectures. As an inheritance from traditional CNNs, most approaches formulate graph pooling as a cluster assignment problem, extending the idea of local patches in regular grids to graphs. Despite the wide adherence to this design choice, no work has rigorously evaluated its influence on the success of GNNs. In this paper, we build upon representative GNNs and introduce variants that challenge the need for locality-preserving representations, either using randomization or clustering on the complement graph. Strikingly, our experiments demonstrate that using these variants does not result in any decrease in performance. To understand this phenomenon, we study the interplay between convolutional layers and the subsequent pooling ones. We show that the convolutions play a leading role in the learned representations. In contrast to the common belief, local pooling is not responsible for the success of GNNs on relevant and widely-used benchmarks.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
30 top-tier papers have cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Open Graph Benchmark: Datasets for Machine Learning on Graphs cites A Fair Comparison of Graph Neural Networks for Graph Classification
- Structure-Aware Transformer for Graph Representation Learning cites Open Graph Benchmark: Datasets for Machine Learning on Graphs
- Structure-Aware Transformer for Graph Representation Learning cites Representing Long-Range Context for Graph Neural Networks with Global Attention
- Structural Entropy Guided Graph Hierarchical Pooling cites A Fair Comparison of Graph Neural Networks for Graph Classification
- Structural Entropy Guided Graph Hierarchical Pooling cites Spectral Clustering with Graph Neural Networks for Graph Pooling
- Representing Long-Range Context for Graph Neural Networks with Global Attention cites Open Graph Benchmark: Datasets for Machine Learning on Graphs
- Representing Long-Range Context for Graph Neural Networks with Global Attention cites AttPool: Towards Hierarchical Feature Representation in Graph Convolutional Networks via Attention Mechanism
- Representing Long-Range Context for Graph Neural Networks with Global Attention cites Memory-Based Graph Networks
- Representing Long-Range Context for Graph Neural Networks with Global Attention cites A Fair Comparison of Graph Neural Networks for Graph Classification
Cited by top-tier papers30
- Representing Long-Range Context for Graph Neural Networks with Global AttentionZhanghao Wu, Paras Jain, Matthew A. Wright, Azalia Mirhoseini et al.NeurIPS 2021 · 450 citations
- Structure-Aware Transformer for Graph Representation LearningDexiong Chen, Leslie O'Bray, Karsten M. BorgwardtICML 2022 · 349 citations
- Universal Prompt Tuning for Graph Neural NetworksTaoran Fang, Yunchao Zhang, Yang Yang, Chunping Wang et al.NeurIPS 2023 · 166 citations
- Mole-BERT: Rethinking Pre-training Graph Neural Networks for MoleculesJun Xia, Chengshuai Zhao, Bozhen Hu, Zhangyang Gao et al.ICLR 2023 · 119 citations
- Structural Entropy Guided Graph Hierarchical PoolingJunran Wu, Xueyuan Chen, Ke Xu, Shangzhe LiICML 2022 · 113 citations
Builds on5
- Open Graph Benchmark: Datasets for Machine Learning on GraphsWeihua Hu, Matthias Fey, Marinka Zitnik, Yuxiao Dong et al.NeurIPS 2020 · 3,935 citations
- Spectral Clustering with Graph Neural Networks for Graph PoolingFilippo Maria Bianchi, Daniele Grattarola, Cesare AlippiICML 2020 · 528 citations
- A Fair Comparison of Graph Neural Networks for Graph ClassificationFederico Errica, Marco Podda, Davide Bacciu, Alessio MicheliICLR 2020 · 508 citations
- Memory-Based Graph NetworksAmir Hosein Khas Ahmadi, Kaveh Hassani, Parsa Moradi, Leo Lee et al.ICLR 2020 · 100 citations
- AttPool: Towards Hierarchical Feature Representation in Graph Convolutional Networks via Attention MechanismJingjia Huang, Zhangheng Li, Nannan Li, Shan Liu et al.ICCV 2019 · 59 citations
Related papers
- Topological Pooling on GraphsYuzhou Chen, Yulia R. GelAAAI 2023 · 21 citations
- Geometry-Aware Edge Pooling for Graph Neural NetworksKatharina Limbeck, Lydia Mezrag, Guy Wolf, Bastian RieckNeurIPS 2025 · 9 citations
- SSHPool: The Separated Subgraph-based Hierarchical PoolingZhuo Xu, Lu Bai, Lixin Cui, Ming Li et al.AAAI 2026 · 1 citation
- The expressive power of pooling in Graph Neural NetworksFilippo Maria Bianchi, Veronica LachiNeurIPS 2023 · 55 citations
- Boosting Graph Pooling with Persistent HomologyChaolong Ying, Xinjian Zhao, Tianshu YuNeurIPS 2024 · 20 citations
