Pooling in Graph Convolutional Neural Networks

  title={Pooling in Graph Convolutional Neural Networks},
  author={Mark Cheung and John Shi and Lavender Yao Jiang and Oren Wright and Jos{\'e} M. F. Moura},
  journal={2019 53rd Asilomar Conference on Signals, Systems, and Computers},
  • Mark CheungJohn Shi J. Moura
  • Published 1 November 2019
  • Computer Science
  • 2019 53rd Asilomar Conference on Signals, Systems, and Computers
Graph convolutional neural networks (GCNNs) are a powerful extension of deep learning techniques to graph-structured data problems. We empirically evaluate several pooling methods for GCNNs, and combinations of those graph pooling methods with three different architectures: GCN, TAGCN, and GraphSAGE. We confirm that graph pooling, especially DiffPool, improves classification accuracy on popular graph classification datasets and find that, on average, TAGCN achieves comparable or better accuracy… 

Figures and Tables from this paper

Graph Signal Processing and Deep Learning: Convolution, Pooling, and Topology

This article explores how graph signal processing can be used to extend CNN components to graphs to improve model performance and how to design the graph CNN architecture based on the topology or structure of the data graph.

Learning Similarity-Preserving Representations of Brain Structure-Function Coupling

  • Yang LiG. Mateos
  • Computer Science
    2022 30th European Signal Processing Conference (EUSIPCO)
  • 2022
A supervised graph representation learning framework to model the relationship between brain structural connectivity (SC) and functional connectivity (FC) via a graph encoder-decoder system equipped with graph convolutional encoders to learn graph-level embeddings that preserve application-dependent similarity measures between brain networks.

Learning to Model the Relationship Between Brain Structural and Functional Connectomes

A supervised graph representation learning framework to model the relationship between brain structural connectivity (SC) and functional connectivity (FC) via a graph encoder-decoder system, where the SC is used as input to predict empirical FC.

Comparison of Malware Classification Methods using Convolutional Neural Network based on API Call Stream

The proposed 1-D CNN outperformed other classification techniques with 91% overall accuracy for both categorical and TFIDF vectors.


A Convolutional Neural Network for malware type classification based on the Windows system API (Application Program Interface) calls is presented and the proposed 1-D CNN outperformed other traditional classification techniques with overall accuracy score of 91.0%.



Hierarchical Graph Representation Learning with Differentiable Pooling

DiffPool is proposed, a differentiable graph pooling module that can generate hierarchical representations of graphs and can be combined with various graph neural network architectures in an end-to-end fashion.

Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering

This work presents a formulation of CNNs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs.

Clique pooling for graph classification

A novel graph pooling operation using cliques as the unit pool is proposed, more readily interpretable, a better analogue to image coarsening than filtering or pruning techniques, and entirely nonparametric.


This paper analyzes these two types of graph convolutional layers and demonstrates that the spectrum domain based graph convo-lutional layer suffers from output inconsistencies when the graph shift matrix has repeated eigenvalues.

Classification with Vertex-Based Graph Convolutional Neural Networks

TAGCN is used to classify different time periods during the week based on New York City taxi data defined on a directed graph with an accuracy of 88% using a single graph convolutional layer.

Semi-Supervised Classification with Graph Convolutional Networks

A scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs which outperforms related methods by a significant margin.

Spectral Networks and Locally Connected Networks on Graphs

This paper considers possible generalizations of CNNs to signals defined on more general domains without the action of a translation group, and proposes two constructions, one based upon a hierarchical clustering of the domain, and another based on the spectrum of the graph Laplacian.

Self-Attention Graph Pooling

This paper proposes a graph pooling method based on self-attention using graph convolution, which achieves superior graph classification performance on the benchmark datasets using a reasonable number of parameters.

Topology adaptive graph convolutional networks

The proposed TAGCN not only inherits the properties of convolutions in CNN for grid-structured data, but it is also consistent with convolution as defined in graph signal processing, exhibiting better performance than existing graph-convolution-approximation methods on a number of data sets.

An End-to-End Deep Learning Architecture for Graph Classification

This paper designs a localized graph convolution model and shows its connection with two graph kernels, and designs a novel SortPooling layer which sorts graph vertices in a consistent order so that traditional neural networks can be trained on the graphs.