Approximate Fully Connected Neural Network Generation

Tuba Ayhan; Mustafa Altun

doi:10.1109/SMACD.2018.8434843

Approximate Fully Connected Neural Network Generation

Tuba Ayhan, Mustafa Altun

Department of Electronics and Communication Engineering

Istanbul Technical University

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

6 Citations (Scopus)

Abstract

Approximate computing is exploited in implementation of fully connected networks for classification problems. A multiplier structure whose area is scalable over accuracy through approximate computing is proposed. In order to employ the multipliers in a network, an area reduction algorithm is formed. It can adjust the approximation level of multipliers while still maintaining the target classification performance, without prior information on the value of network weights. Implementing on a Spartan6 FPGA, up to 79% area saving is recorded for various performance targets.

Original language	English
Title of host publication	SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	93-96
Number of pages	4
ISBN (Print)	9781538651520
DOIs	https://doi.org/10.1109/SMACD.2018.8434843
Publication status	Published - 13 Aug 2018
Event	15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, SMACD 2018 - Prague, Czech Republic Duration: 2 Jul 2018 → 5 Jul 2018

Publication series

Name	SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design

Conference

Conference	15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, SMACD 2018
Country/Territory	Czech Republic
City	Prague
Period	2/07/18 → 5/07/18

Bibliographical note

Publisher Copyright:
© 2018 IEEE.

Funding

ACKNOWLEDGEMENTS This project is funded by TUBITAK (The Scientific and Technological Research Council of Turkey), with the grant number 117E078.

Funders	Funder number
TUBITAK
Türkiye Bilimsel ve Teknolojik Araştirma Kurumu	117E078

Keywords

Approximate computing
area reduction
fully connected network

Access to Document

10.1109/SMACD.2018.8434843

Cite this

Ayhan, T., & Altun, M. (2018). Approximate Fully Connected Neural Network Generation. In SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design (pp. 93-96). Article 8434843 (SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SMACD.2018.8434843

Ayhan, Tuba ; Altun, Mustafa. / Approximate Fully Connected Neural Network Generation. SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 93-96 (SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design).

@inproceedings{14875496cbbe4b25ae1f4232e901912d,

title = "Approximate Fully Connected Neural Network Generation",

abstract = "Approximate computing is exploited in implementation of fully connected networks for classification problems. A multiplier structure whose area is scalable over accuracy through approximate computing is proposed. In order to employ the multipliers in a network, an area reduction algorithm is formed. It can adjust the approximation level of multipliers while still maintaining the target classification performance, without prior information on the value of network weights. Implementing on a Spartan6 FPGA, up to 79% area saving is recorded for various performance targets.",

keywords = "Approximate computing, area reduction, fully connected network",

author = "Tuba Ayhan and Mustafa Altun",

note = "Publisher Copyright: {\textcopyright} 2018 IEEE.; 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, SMACD 2018 ; Conference date: 02-07-2018 Through 05-07-2018",

year = "2018",

month = aug,

day = "13",

doi = "10.1109/SMACD.2018.8434843",

language = "English",

isbn = "9781538651520",

series = "SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design",

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

pages = "93--96",

booktitle = "SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design",

address = "United States",

}

Ayhan, T & Altun, M 2018, Approximate Fully Connected Neural Network Generation. in SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design., 8434843, SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, Institute of Electrical and Electronics Engineers Inc., pp. 93-96, 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, SMACD 2018, Prague, Czech Republic, 2/07/18. https://doi.org/10.1109/SMACD.2018.8434843

Approximate Fully Connected Neural Network Generation. / Ayhan, Tuba; Altun, Mustafa.
SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design. Institute of Electrical and Electronics Engineers Inc., 2018. p. 93-96 8434843 (SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design).

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

TY - GEN

T1 - Approximate Fully Connected Neural Network Generation

AU - Ayhan, Tuba

AU - Altun, Mustafa

PY - 2018/8/13

Y1 - 2018/8/13

N2 - Approximate computing is exploited in implementation of fully connected networks for classification problems. A multiplier structure whose area is scalable over accuracy through approximate computing is proposed. In order to employ the multipliers in a network, an area reduction algorithm is formed. It can adjust the approximation level of multipliers while still maintaining the target classification performance, without prior information on the value of network weights. Implementing on a Spartan6 FPGA, up to 79% area saving is recorded for various performance targets.

AB - Approximate computing is exploited in implementation of fully connected networks for classification problems. A multiplier structure whose area is scalable over accuracy through approximate computing is proposed. In order to employ the multipliers in a network, an area reduction algorithm is formed. It can adjust the approximation level of multipliers while still maintaining the target classification performance, without prior information on the value of network weights. Implementing on a Spartan6 FPGA, up to 79% area saving is recorded for various performance targets.

KW - Approximate computing

KW - area reduction

KW - fully connected network

UR - http://www.scopus.com/inward/record.url?scp=85052540696&partnerID=8YFLogxK

U2 - 10.1109/SMACD.2018.8434843

DO - 10.1109/SMACD.2018.8434843

M3 - Conference contribution

AN - SCOPUS:85052540696

SN - 9781538651520

T3 - SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design

SP - 93

EP - 96

BT - SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design, SMACD 2018

Y2 - 2 July 2018 through 5 July 2018

ER -

Ayhan T, Altun M. Approximate Fully Connected Neural Network Generation. In SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design. Institute of Electrical and Electronics Engineers Inc. 2018. p. 93-96. 8434843. (SMACD 2018 - 15th International Conference on Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit Design). doi: 10.1109/SMACD.2018.8434843

Approximate Fully Connected Neural Network Generation

Abstract

Publication series

Conference

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this