SpaPool: Soft Partition Assignment Pooling for__Graph Neural Networks

TL;DR

SpaPool tackles graph pooling for GNNs by marrying dense clustering with adaptive node selection to preserve both local and global graph structure while reducing size. It defines a soft assignment matrix S to align representative centroids with node embeddings and reduces the graph by updating node features and adjacency via S, aided by an auxiliary loss inspired by DiffPool. Evaluations on ten graph datasets show SpaPool delivering competitive accuracy relative to existing pooling methods, with notable gains on small graphs where dense clustering helps retain information. The work points to SpaPool as a promising, scalable pooling option and suggests future work on scaling to large graphs and adding explainability.

Abstract

This paper introduces SpaPool, a novel pooling method that combines the strengths of both dense and sparse techniques for a graph neural network. SpaPool groups vertices into an adaptive number of clusters, leveraging the benefits of both dense and sparse approaches. It aims to maintain the structural integrity of the graph while reducing its size efficiently. Experimental results on several datasets demonstrate that SpaPool achieves competitive performance compared to existing pooling techniques and excels particularly on small-scale graphs. This makes SpaPool a promising method for applications requiring efficient and effective graph processing.

Figures (2)

• Figure 1: Illustration of the proposed graph pooling method (SpaPool).
• Figure 2: Graph Neural Network model used in the experiments.