Graph Reasoning Networks

TL;DR

Graph Reasoning Networks (GRNs) enhance graph neural networks by integrating fixed and learned representations with a differentiable reasoning module, improving high-level reasoning capabilities demonstrated on synthetic datasets.

Contribution

The paper introduces GRNs, a novel framework combining fixed and learned graph representations with a differentiable satisfiability solver for improved reasoning.

Findings

Comparable performance to GNNs on real-world data

Superior reasoning ability on synthetic datasets

Potential for high-level reasoning enhancement

Abstract

Graph neural networks (GNNs) are the predominant approach for graph-based machine learning. While neural networks have shown great performance at learning useful representations, they are often criticized for their limited high-level reasoning abilities. In this work, we present Graph Reasoning Networks (GRNs), a novel approach to combine the strengths of fixed and learned graph representations and a reasoning module based on a differentiable satisfiability solver. While results on real-world datasets show comparable performance to GNN, experiments on synthetic datasets demonstrate the potential of the newly proposed method.