Configuring Safe Spiking Neural Controllers for Cyber-Physical Systems through Formal Verification

Document Type

Conference Article

Publication Title

Proceedings 2024 22nd ACM IEEE International Symposium on Formal Methods and Models for System Design Memocode 2024

Abstract

In this paper, we address the problem of safety verification for Spiking Neural Networks (SNNs) with Spiking Rectified Linear Activation (SRLA). The SNNs are obtained by first training Artificial Neural Networks (ANNs) and then translating to SNN with subsequent hyperparameter tuning. We propose a solution which tunes the temporal window hyperparameter of the translated SNN to ensure both accuracy and compliance with the safe range specification that requires the SNN outputs to remain within a safe range. We demonstrate our approach with experiments on 5 benchmark neural controllers.

First Page

103

Last Page

107

DOI

10.1109/MEMOCODE63347.2024.00017

Publication Date

1-1-2024

Share

COinS