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

Carlo Abate, Filippo Maria Bianchi

Published 2024-09-08Version 1

We propose a novel approach to compute the \texttt{MAXCUT} in attributed graphs, \textit{i.e.}, graphs with features associated with nodes and edges. Our approach is robust to the underlying graph topology and is fully differentiable, making it possible to find solutions that jointly optimize the \texttt{MAXCUT} along with other objectives. Based on the obtained \texttt{MAXCUT} partition, we implement a hierarchical graph pooling layer for Graph Neural Networks, which is sparse, differentiable, and particularly suitable for downstream tasks on heterophilic graphs.