Dependable Neural Networks Through Redundancy, A Comparison of Redundant Architectures

TL;DR

This paper investigates the dependability of neural networks in edge-AI applications, comparing redundant architectures and exploring synchronization challenges to enhance safe and reliable AI deployment.

Contribution

It provides a comparative analysis of redundant neural network architectures and discusses implementation challenges for dependable AI systems.

Findings

Preliminary measurements support the need for synchronization in redundant neural networks.

Work on implementing lockstep neural network engines is introduced.

Insights into the reliability of redundant architectures for safety-critical applications.

Abstract

With edge-AI finding an increasing number of real-world applications, especially in industry, the question of functionally safe applications using AI has begun to be asked. In this body of work, we explore the issue of achieving dependable operation of neural networks. We discuss the issue of dependability in general implementation terms before examining lockstep solutions. We intuit that it is not necessarily a given that two similar neural networks generate results at precisely the same time and that synchronization between the platforms will be required. We perform some preliminary measurements that may support this intuition and introduce some work in implementing lockstep neural network engines.