Rethinking pooling in graph neural networks

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.

arXiv commentAccepted to NeurIPS 2020

References in corpus (17)

• Inductive Representation Learning on Large Graphs
• Neural Message Passing for Quantum Chemistry
• Geometric deep learning: going beyond Euclidean data
• Fast Graph Representation Learning with PyTorch Geometric
• Simplifying Graph Convolutional Networks
• Junction Tree Variational Autoencoder for Molecular Graph Generation
• Pitfalls of Graph Neural Network Evaluation
• TUDataset: A collection of benchmark datasets for learning with graphs
• Revisiting Graph Neural Networks: All We Have is Low-Pass Filters
• Benchmarking Graph Neural Networks
• A Fair Comparison of Graph Neural Networks for Graph Classification
• Are Powerful Graph Neural Nets Necessary? A Dissection on Graph Classification
• A simple yet effective baseline for non-attributed graph classification
• Pooling is neither necessary nor sufficient for appropriate deformation stability in CNNs
• Understanding Isomorphism Bias in Graph Data Sets
• Memory-Based Graph Networks
• EasyConvPooling: Random Pooling with Easy Convolution for Accelerating Training and Testing

Cited by in corpus (14)

• Understanding Pooling in Graph Neural Networks
• Graph Rationalization with Environment-based Augmentations
• Accurate Learning of Graph Representations with Graph Multiset Pooling
• On Representation Knowledge Distillation for Graph Neural Networks
• Higher-order Clustering and Pooling for Graph Neural Networks
• A Novel Normalized-Cut Solver with Nearest Neighbor Hierarchical Initialization
• Search to Capture Long-range Dependency with Stacking GNNs for Graph Classification
• Spotting Virus from Satellites: Modeling the Circulation of West Nile Virus Through Graph Neural Networks
• On Positional and Structural Node Features for Graph Neural Networks on Non-attributed Graphs
• Accelerating Backward Aggregation in GCN Training with Execution Path Preparing on GPUs
• Evaluating Modules in Graph Contrastive Learning
• NeuroBack: Improving CDCL SAT Solving using Graph Neural Networks
• ChebLieNet: Invariant Spectral Graph NNs Turned Equivariant by Riemannian Geometry on Lie Groups
• Diversified Multiscale Graph Learning with Graph Self-Correction