Uncertainty Modeling of Emerging Device-based Computing-in-Memory Neural Accelerators with Application to Neural Architecture Search

TL;DR

This paper analyzes the impact of device uncertainties in emerging computing-in-memory neural accelerators and introduces UAE, a neural architecture search method that finds models resilient to these uncertainties.

Contribution

The paper provides a detailed analysis of device uncertainty effects and proposes UAE, a novel uncertainty-aware neural architecture search scheme for robust DNN models.

Findings

Uncertainty significantly affects DNN accuracy on emerging devices.

UAE effectively identifies models with improved robustness against uncertainties.

The approach enhances energy efficiency without sacrificing accuracy.

Abstract

Emerging device-based Computing-in-memory (CiM) has been proved to be a promising candidate for high-energy efficiency deep neural network (DNN) computations. However, most emerging devices suffer uncertainty issues, resulting in a difference between actual data stored and the weight value it is designed to be. This leads to an accuracy drop from trained models to actually deployed platforms. In this work, we offer a thorough analysis of the effect of such uncertainties-induced changes in DNN models. To reduce the impact of device uncertainties, we propose UAE, an uncertainty-aware Neural Architecture Search scheme to identify a DNN model that is both accurate and robust against device uncertainties.