Analyzing B\"uchi Automata with Graph Neural Networks

TL;DR

This paper explores using graph neural networks to analyze B"uchi automata, aiming to improve efficiency in solving computationally hard problems in program verification by leveraging graph representations.

Contribution

It introduces a novel approach of applying graph neural networks to predict properties of B"uchi automata, demonstrating effectiveness on automatically generated datasets.

Findings

GNNs can reliably predict properties of B"uchi automata

The approach outperforms traditional methods in certain tasks

Automated dataset generation facilitates training

Abstract

B\"uchi Automata on infinite words present many interesting problems and are used frequently in program verification and model checking. A lot of these problems on B\"uchi automata are computationally hard, raising the question if a learning-based data-driven analysis might be more efficient than using traditional algorithms. Since B\"uchi automata can be represented by graphs, graph neural networks are a natural choice for such a learning-based analysis. In this paper, we demonstrate how graph neural networks can be used to reliably predict basic properties of B\"uchi automata when trained on automatically generated random automata datasets.