Implementing a CTL Model Checker with $$\mu \mathcal {G}$$μG, a Language for Programming Graph Neural Networks - Formal Techniques for Distributed Objects, Components, and Systems
Conference Papers Year : 2023

Implementing a CTL Model Checker with $$\mu \mathcal {G}$$μG, a Language for Programming Graph Neural Networks

Abstract

A graph neural network is a deep learning architecture operating on graph-structured data. While they have achieved impressive results in many application domains, their applicability is limited when one needs to perform symbolic and semantic reasoning, for example, in model checking. We propose an approach based on the idea that graph neural networks can be programmed. We introduce $$\mu \mathcal {G}$$μG, a simple language inspired by modal $$\mu $$μ-calculus that can be compiled, via TensorFlow, in a graph neural network that can be efficiently run on parallel architectures. To demonstrate our method, we use our language to implement a Computation Tree Logic model checker. A benchmark from the Model Checking Contest is used to show that the performance of the resulting tool outperforms that of other model checkers.
Embargoed file
Embargoed file
1 2 4
Year Month Jours
Avant la publication
Thursday, January 1, 2026
Embargoed file
Thursday, January 1, 2026
Please log in to request access to the document

Dates and versions

hal-04731936 , version 1 (11-10-2024)

Licence

Identifiers

Cite

Matteo Belenchia, Flavio Corradini, Michela Quadrini, Michele Loreti. Implementing a CTL Model Checker with $$\mu \mathcal {G}$$μG, a Language for Programming Graph Neural Networks. 43th International Conference on Formal Techniques for Distributed Objects, Components, and Systems (FORTE), Jun 2023, Lisbon, Portugal. pp.37-54, ⟨10.1007/978-3-031-35355-0_4⟩. ⟨hal-04731936⟩
0 View
0 Download

Altmetric

Share

More