Hybrid Convolutional Neural Networks with Reliability Guarantee

TL;DR

This paper introduces a hybrid CNN approach that combines reliable and non-reliable execution to enhance AI safety and dependability while minimizing additional computational costs.

Contribution

It presents a novel co-design method integrating redundant reliable execution selectively within CNNs to improve dependability efficiently.

Findings

Preliminary results demonstrate reduced computational overhead.

Hybrid CNN achieves dependable execution with minimal extra cost.

The approach extends AI-accelerator safety and dependability capabilities.

Abstract

Making AI safe and dependable requires the generation of dependable models and dependable execution of those models. We propose redundant execution as a well-known technique that can be used to ensure reliable execution of the AI model. This generic technique will extend the application scope of AI-accelerators that do not feature well-documented safety or dependability properties. Typical redundancy techniques incur at least double or triple the computational expense of the original. We adopt a co-design approach, integrating reliable model execution with non-reliable execution, focusing that additional computational expense only where it is strictly necessary. We describe the design, implementation and some preliminary results of a hybrid CNN.