Cargando…

Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge

Previous studies have demonstrated that, up to a certain degree, Convolutional Neural Networks (CNNs) can tolerate arithmetic approximations. Nonetheless, perturbations must be applied judiciously, to constrain their impact on accuracy. This is a challenging task, since the implementation of inexact...

Descripción completa

Detalles Bibliográficos
Autores principales: Ponzina, Flavio, Ansaloni, Giovanni, Peón-Quirós, Miguel, Atienza, David
Formato: Online Artículo Texto
Lenguaje:English
Publicado: MDPI 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9320279/
https://www.ncbi.nlm.nih.gov/pubmed/35888960
http://dx.doi.org/10.3390/mi13071143
_version_ 1784755754076471296
author Ponzina, Flavio
Ansaloni, Giovanni
Peón-Quirós, Miguel
Atienza, David
author_facet Ponzina, Flavio
Ansaloni, Giovanni
Peón-Quirós, Miguel
Atienza, David
author_sort Ponzina, Flavio
collection PubMed
description Previous studies have demonstrated that, up to a certain degree, Convolutional Neural Networks (CNNs) can tolerate arithmetic approximations. Nonetheless, perturbations must be applied judiciously, to constrain their impact on accuracy. This is a challenging task, since the implementation of inexact operators is often decided at design time, when the application and its robustness profile are unknown, posing the risk of over-constraining or over-provisioning the hardware. Bridging this gap, we propose a two-phase strategy. Our framework first optimizes the target CNN model, reducing the bitwidth of weights and activations and enhancing error resiliency, so that inexact operations can be performed as frequently as possible. Then, it selectively assigns CNN layers to exact or inexact hardware based on a sensitivity metric. Our results show that, within a 5% accuracy degradation, our methodology, including a highly inexact multiplier design, can reduce the cost of MAC operations in CNN inference up to 83.6% compared to state-of-the-art optimized exact implementations.
format Online
Article
Text
id pubmed-9320279
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher MDPI
record_format MEDLINE/PubMed
spelling pubmed-93202792022-07-27 Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge Ponzina, Flavio Ansaloni, Giovanni Peón-Quirós, Miguel Atienza, David Micromachines (Basel) Article Previous studies have demonstrated that, up to a certain degree, Convolutional Neural Networks (CNNs) can tolerate arithmetic approximations. Nonetheless, perturbations must be applied judiciously, to constrain their impact on accuracy. This is a challenging task, since the implementation of inexact operators is often decided at design time, when the application and its robustness profile are unknown, posing the risk of over-constraining or over-provisioning the hardware. Bridging this gap, we propose a two-phase strategy. Our framework first optimizes the target CNN model, reducing the bitwidth of weights and activations and enhancing error resiliency, so that inexact operations can be performed as frequently as possible. Then, it selectively assigns CNN layers to exact or inexact hardware based on a sensitivity metric. Our results show that, within a 5% accuracy degradation, our methodology, including a highly inexact multiplier design, can reduce the cost of MAC operations in CNN inference up to 83.6% compared to state-of-the-art optimized exact implementations. MDPI 2022-07-19 /pmc/articles/PMC9320279/ /pubmed/35888960 http://dx.doi.org/10.3390/mi13071143 Text en © 2022 by the authors. https://creativecommons.org/licenses/by/4.0/Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
spellingShingle Article
Ponzina, Flavio
Ansaloni, Giovanni
Peón-Quirós, Miguel
Atienza, David
Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge
title Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge
title_full Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge
title_fullStr Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge
title_full_unstemmed Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge
title_short Using Algorithmic Transformations and Sensitivity Analysis to Unleash Approximations in CNNs at the Edge
title_sort using algorithmic transformations and sensitivity analysis to unleash approximations in cnns at the edge
topic Article
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9320279/
https://www.ncbi.nlm.nih.gov/pubmed/35888960
http://dx.doi.org/10.3390/mi13071143
work_keys_str_mv AT ponzinaflavio usingalgorithmictransformationsandsensitivityanalysistounleashapproximationsincnnsattheedge
AT ansalonigiovanni usingalgorithmictransformationsandsensitivityanalysistounleashapproximationsincnnsattheedge
AT peonquirosmiguel usingalgorithmictransformationsandsensitivityanalysistounleashapproximationsincnnsattheedge
AT atienzadavid usingalgorithmictransformationsandsensitivityanalysistounleashapproximationsincnnsattheedge