August 11, 2025

Towards Logical Representations of Recurrent Neural Networks

Key Points

Recurrent neural networks can be represented logically, improving their interpretability and facilitating verification.
Rational McNaughton functions are employed to enhance representations, allowing for a clearer understanding of neural behavior.
The proposed method focuses on simple recurrent networks, marking a step toward broader applications in deep learning.
Logical frameworks like Łukasiewicz logic are central to this representation, showcasing a novel intersection between neural networks and formal logic.

Abstract

Representations of neural networks in the formal language of logical systems may be used to enhance their interpretability and as a step in the formal verification of their properties. This work builds on the literature regarding the representation of feedforward neural networks that compute rational McNaughton functions in the language of Łukasiewicz logic. Thus, we propose a technique for representing recurrent computations, enabling the representation of recurrent neural networks. This is an initial investigation conducted within the specific context of simple recurrent networks (SRNs) whose layers compute rational McNaughton functions.

Mark Helpful

Bookmark

Relay

Cite This Study

Sandro Preto (Thu,) studied this question.

synapsesocial.com/papers/68a35ee30a429f7973327bbb https://doi.org/https://doi.org/10.5753/wbl.2025.6543

Mark Helpful

Bookmark

Relay