Logic Gate Neural Networks are Good for Verification

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Kresse, Fabian, Yu, Emily, Lampert, Christoph H., Henzinger, Thomas A.
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911180560269312
author Kresse, Fabian
Yu, Emily
Lampert, Christoph H.
Henzinger, Thomas A.
author_facet Kresse, Fabian
Yu, Emily
Lampert, Christoph H.
Henzinger, Thomas A.
contents Learning-based systems are increasingly deployed across various domains, yet the complexity of traditional neural networks poses significant challenges for formal verification. Unlike conventional neural networks, learned Logic Gate Networks (LGNs) replace multiplications with Boolean logic gates, yielding a sparse, netlist-like architecture that is inherently more amenable to symbolic verification, while still delivering promising performance. In this paper, we introduce a SAT encoding for verifying global robustness and fairness in LGNs. We evaluate our method on five benchmark datasets, including a newly constructed 5-class variant, and find that LGNs are both verification-friendly and maintain strong predictive performance.
format Preprint
id arxiv_https___arxiv_org_abs_2505_19932
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Logic Gate Neural Networks are Good for Verification
Kresse, Fabian
Yu, Emily
Lampert, Christoph H.
Henzinger, Thomas A.
Machine Learning
Logic in Computer Science
Learning-based systems are increasingly deployed across various domains, yet the complexity of traditional neural networks poses significant challenges for formal verification. Unlike conventional neural networks, learned Logic Gate Networks (LGNs) replace multiplications with Boolean logic gates, yielding a sparse, netlist-like architecture that is inherently more amenable to symbolic verification, while still delivering promising performance. In this paper, we introduce a SAT encoding for verifying global robustness and fairness in LGNs. We evaluate our method on five benchmark datasets, including a newly constructed 5-class variant, and find that LGNs are both verification-friendly and maintain strong predictive performance.
title Logic Gate Neural Networks are Good for Verification
topic Machine Learning
Logic in Computer Science
url https://arxiv.org/abs/2505.19932