Home // INTELLI 2023, The Twelfth International Conference on Intelligent Systems and Applications // View article

A Novel Diagnosis Concept for Verification of Neural Networks on the Target Hardware

Authors:
Ilkay Wunderlich
Jens Braunes
Rainer G. Spallek

Keywords: neural networks, hardware, diagnosis, debugging.

Abstract:
Neural networks are increasingly being used in embedded applications. It is important to check whether the trained network fulfills its tasks not only in simulations but also on real target hardware. This is of particular importance in safety-critical applications such as plant control or autonomous driving. For this reason, a diagnostic concept for AI-based systems based on the Universal Debug Engine from PLS Programmierbare Logik & Systeme GmbH was developed at the Technical University of Dresden. This concept enables developers to verify the hardware implementation of the neural network. After the concept was developed, it was successfully tested on several example applications of which two are presented in the paper.

Pages: 1 to 6

Copyright: Copyright (c) IARIA, 2023

Publication date: March 13, 2023

Published in: conference

ISSN: 2308-4065

ISBN: 978-1-68558-064-3

Location: Barcelona, Spain

Dates: from March 13, 2023 to March 17, 2023