MaxCutPool: differentiable feature-aware Maxcut for pooling in graph neural networks

TL;DR

This paper introduces MaxCutPool, a differentiable, feature-aware MAXCUT method for graph pooling in neural networks, enabling end-to-end training and improved performance on heterophilic graphs.

Contribution

It presents a novel MAXCUT computation method that integrates with GNNs for hierarchical pooling, adaptable to various graph topologies and optimized jointly with other objectives.

Findings

Effective on attributed graphs with diverse topologies

Enables end-to-end training of graph pooling layers

Improves downstream task performance on heterophilic graphs

Abstract

We propose a novel approach to compute the MAXCUT in attributed graphs, i.e., graphs with features associated with nodes and edges. Our approach works well on any kind of graph topology and can find solutions that jointly optimize the MAXCUT along with other objectives. Based on the obtained MAXCUT partition, we implement a hierarchical graph pooling layer for Graph Neural Networks, which is sparse, trainable end-to-end, and particularly suitable for downstream tasks on heterophilic graphs.