Graph neural networks to learn joint representations of disjoint molecular graphs

TL;DR

This paper explores the challenge of learning global representations for disjoint molecular graphs using graph neural networks, introducing a new dataset and analyzing initial results that highlight generalization difficulties.

Contribution

It introduces a new dataset of disjoint molecular graphs with joint labels and evaluates graph neural networks on this task, revealing current limitations.

Findings

GNNs can solve the task on a subset of the dataset

Models struggle to generalize to unseen (sub)graphs

The dataset highlights challenges in representing disjoint graphs

Abstract

Graph neural networks are widely used to learn global representations of graphs, which are then used for regression or classification tasks. Typically, the graphs in such data sets are connected, i.e. each training sample consists of a single internally connected graph associated with a global label. However, there is a wide variety of yet unconsidered but application-relevant tasks, where labels are assigned to sets of disjoint graphs, which requires the generation of global representations of disjoint graphs. In this paper, we present a new data set with chemical reactions, which is illustrating this task. Each sample consists of a pair of disjoint molecular graphs and a joint label representing a scalar measure associated with the chemical reaction of the molecules. We show the initial results of graph neural networks that are able to solve the task within a combinatorial subset of the dataset but do not generalize well to the full data set and unseen (sub)graphs.