Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations

Alberto Marchisio; Beatrice Bussolino; Edoardo Salvati; Maurizio Martina; Guido Masera; Muhammad Shafique

doi:10.1145/3531437.3539717

Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations

Alberto Marchisio, Beatrice Bussolino, Edoardo Salvati, Maurizio Martina, Guido Masera, Muhammad Shafique

Electrical Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Complex Deep Neural Networks such as Capsule Networks (CapsNets) exhibit high learning capabilities at the cost of computeintensive operations. To enable their deployment on edge devices, we propose to leverage approximate computing for designing approximate variants of the complex operations like softmax and squash. In our experiments, we evaluate tradeoffs between area, power consumption, and critical path delay of the designs implemented with the ASIC design flow, and the accuracy of the quantized CapsNets, compared to the exact functions.

Original language	English (US)
Title of host publication	2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781450393546
DOIs	https://doi.org/10.1145/3531437.3539717
State	Published - Aug 2 2022
Event	2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022 - Virtual, Online, United States Duration: Aug 1 2022 → Aug 2 2022

Publication series

Name	Proceedings of the International Symposium on Low Power Electronics and Design
ISSN (Print)	1533-4678

Conference

Conference	2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022
Country/Territory	United States
City	Virtual, Online
Period	8/1/22 → 8/2/22

Keywords

Approximate Computing
Capsule Networks
Deep Neural Networks
Nonlinear Functions
Softmax
Squash

ASJC Scopus subject areas

General Engineering

Access to Document

10.1145/3531437.3539717

Cite this

Marchisio, A., Bussolino, B., Salvati, E., Martina, M., Masera, G., & Shafique, M. (2022). Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations. In 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022 Article 27 (Proceedings of the International Symposium on Low Power Electronics and Design). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1145/3531437.3539717

Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations. / Marchisio, Alberto; Bussolino, Beatrice; Salvati, Edoardo et al.
2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022. Institute of Electrical and Electronics Engineers Inc., 2022. 27 (Proceedings of the International Symposium on Low Power Electronics and Design).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Marchisio, A, Bussolino, B, Salvati, E, Martina, M, Masera, G & Shafique, M 2022, Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations. in 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022., 27, Proceedings of the International Symposium on Low Power Electronics and Design, Institute of Electrical and Electronics Engineers Inc., 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022, Virtual, Online, United States, 8/1/22. https://doi.org/10.1145/3531437.3539717

Marchisio A, Bussolino B, Salvati E, Martina M, Masera G, Shafique M. Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations. In 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022. Institute of Electrical and Electronics Engineers Inc. 2022. 27. (Proceedings of the International Symposium on Low Power Electronics and Design). doi: 10.1145/3531437.3539717

Marchisio, Alberto ; Bussolino, Beatrice ; Salvati, Edoardo et al. / Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations. 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022. Institute of Electrical and Electronics Engineers Inc., 2022. (Proceedings of the International Symposium on Low Power Electronics and Design).

@inproceedings{9db002aeca1948a89d606e546ad84b24,

title = "Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations",

abstract = "Complex Deep Neural Networks such as Capsule Networks (CapsNets) exhibit high learning capabilities at the cost of computeintensive operations. To enable their deployment on edge devices, we propose to leverage approximate computing for designing approximate variants of the complex operations like softmax and squash. In our experiments, we evaluate tradeoffs between area, power consumption, and critical path delay of the designs implemented with the ASIC design flow, and the accuracy of the quantized CapsNets, compared to the exact functions.",

keywords = "Approximate Computing, Capsule Networks, Deep Neural Networks, Nonlinear Functions, Softmax, Squash",

author = "Alberto Marchisio and Beatrice Bussolino and Edoardo Salvati and Maurizio Martina and Guido Masera and Muhammad Shafique",

note = "Funding Information: This work has been supported in part by the Doctoral College Resilient Embedded Systems, which is run jointly by the TU Wien{\textquoteright}s Faculty of Informatics and the UAS Technikum Wien. Publisher Copyright: {\textcopyright} 2022 Copyright held by the owner/author(s).; 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022 ; Conference date: 01-08-2022 Through 02-08-2022",

year = "2022",

month = aug,

day = "2",

doi = "10.1145/3531437.3539717",

language = "English (US)",

series = "Proceedings of the International Symposium on Low Power Electronics and Design",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

booktitle = "2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022",

}

TY - GEN

T1 - Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations

AU - Marchisio, Alberto

AU - Bussolino, Beatrice

AU - Salvati, Edoardo

AU - Martina, Maurizio

AU - Masera, Guido

AU - Shafique, Muhammad

N1 - Funding Information: This work has been supported in part by the Doctoral College Resilient Embedded Systems, which is run jointly by the TU Wien’s Faculty of Informatics and the UAS Technikum Wien. Publisher Copyright: © 2022 Copyright held by the owner/author(s).

PY - 2022/8/2

Y1 - 2022/8/2

N2 - Complex Deep Neural Networks such as Capsule Networks (CapsNets) exhibit high learning capabilities at the cost of computeintensive operations. To enable their deployment on edge devices, we propose to leverage approximate computing for designing approximate variants of the complex operations like softmax and squash. In our experiments, we evaluate tradeoffs between area, power consumption, and critical path delay of the designs implemented with the ASIC design flow, and the accuracy of the quantized CapsNets, compared to the exact functions.

AB - Complex Deep Neural Networks such as Capsule Networks (CapsNets) exhibit high learning capabilities at the cost of computeintensive operations. To enable their deployment on edge devices, we propose to leverage approximate computing for designing approximate variants of the complex operations like softmax and squash. In our experiments, we evaluate tradeoffs between area, power consumption, and critical path delay of the designs implemented with the ASIC design flow, and the accuracy of the quantized CapsNets, compared to the exact functions.

KW - Approximate Computing

KW - Capsule Networks

KW - Deep Neural Networks

KW - Nonlinear Functions

KW - Softmax

KW - Squash

UR - http://www.scopus.com/inward/record.url?scp=85136308792&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85136308792&partnerID=8YFLogxK

U2 - 10.1145/3531437.3539717

DO - 10.1145/3531437.3539717

M3 - Conference contribution

AN - SCOPUS:85136308792

T3 - Proceedings of the International Symposium on Low Power Electronics and Design

BT - 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2022 ACM/IEEE International Symposium on Low Power Electronics and Design, ISLPED 2022

Y2 - 1 August 2022 through 2 August 2022

ER -

Enabling Capsule Networks at the Edge through Approximate Softmax and Squash Operations

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this