Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions

Hyun Woo Oh; Won Sik Jeong; Seung Eun Lee

doi:10.1109/ISOCC56007.2022.10031524

Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions

Hyun Woo Oh, Won Sik Jeong, Seung Eun Lee

Electronic Engineering Program

Seoul National University of Science and Technology (SNUST)

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

3 Scopus citations

Abstract

Recent advances in semiconductor technology lead to ongoing applications to adopt complex techniques based on neural networks. In line with this trend, the concept of optimizing real number arithmetic has been raised. In this paper, we evaluate the performance of the noble number system named posit on neural networks by analyzing the execution of approximate exponential functions, which is fundamental to several activation functions, with posit32 and float32. To implement the functions with posit arithmetic, we designed the software posit library consisting of basic arithmetic operations and conversion operations from/to C standard data types. The result shows that posit arithmetic reduces the average relative error rate by up to 87.12% on the exponential function.

Original language	English
Title of host publication	Proceedings - International SoC Design Conference 2022, ISOCC 2022
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	358-359
Number of pages	2
ISBN (Electronic)	9781665459716
DOIs	https://doi.org/10.1109/ISOCC56007.2022.10031524
State	Published - 2022
Event	19th International System-on-Chip Design Conference, ISOCC 2022 - Gangneung-si, Korea, Republic of Duration: 19 Oct 2022 → 22 Oct 2022

Publication series

Name	Proceedings - International SoC Design Conference 2022, ISOCC 2022

Conference

Conference	19th International System-on-Chip Design Conference, ISOCC 2022
Country/Territory	Korea, Republic of
City	Gangneung-si
Period	19/10/22 → 22/10/22

Keywords

activation functions
exponential approximation
IEEE-754
posit
real number arithmetic

Access to Document

10.1109/ISOCC56007.2022.10031524

Cite this

Oh, H. W., Jeong, W. S., & Lee, S. E. (2022). Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions. In Proceedings - International SoC Design Conference 2022, ISOCC 2022 (pp. 358-359). (Proceedings - International SoC Design Conference 2022, ISOCC 2022). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ISOCC56007.2022.10031524

@inproceedings{b89b723134a646698b8d8ec280f9b599,

title = "Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions",

abstract = "Recent advances in semiconductor technology lead to ongoing applications to adopt complex techniques based on neural networks. In line with this trend, the concept of optimizing real number arithmetic has been raised. In this paper, we evaluate the performance of the noble number system named posit on neural networks by analyzing the execution of approximate exponential functions, which is fundamental to several activation functions, with posit32 and float32. To implement the functions with posit arithmetic, we designed the software posit library consisting of basic arithmetic operations and conversion operations from/to C standard data types. The result shows that posit arithmetic reduces the average relative error rate by up to 87.12\% on the exponential function.",

keywords = "activation functions, exponential approximation, IEEE-754, posit, real number arithmetic",

author = "Oh, \{Hyun Woo\} and Jeong, \{Won Sik\} and Lee, \{Seung Eun\}",

note = "Publisher Copyright: {\textcopyright} 2022 IEEE.; 19th International System-on-Chip Design Conference, ISOCC 2022 ; Conference date: 19-10-2022 Through 22-10-2022",

year = "2022",

doi = "10.1109/ISOCC56007.2022.10031524",

language = "English",

series = "Proceedings - International SoC Design Conference 2022, ISOCC 2022",

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

pages = "358--359",

booktitle = "Proceedings - International SoC Design Conference 2022, ISOCC 2022",

}

Oh, HW, Jeong, WS & Lee, SE 2022, Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions. in Proceedings - International SoC Design Conference 2022, ISOCC 2022. Proceedings - International SoC Design Conference 2022, ISOCC 2022, Institute of Electrical and Electronics Engineers Inc., pp. 358-359, 19th International System-on-Chip Design Conference, ISOCC 2022, Gangneung-si, Korea, Republic of, 19/10/22. https://doi.org/10.1109/ISOCC56007.2022.10031524

Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions. / Oh, Hyun Woo; Jeong, Won Sik; Lee, Seung Eun.
Proceedings - International SoC Design Conference 2022, ISOCC 2022. Institute of Electrical and Electronics Engineers Inc., 2022. p. 358-359 (Proceedings - International SoC Design Conference 2022, ISOCC 2022).

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

TY - GEN

T1 - Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions

AU - Oh, Hyun Woo

AU - Jeong, Won Sik

AU - Lee, Seung Eun

PY - 2022

Y1 - 2022

N2 - Recent advances in semiconductor technology lead to ongoing applications to adopt complex techniques based on neural networks. In line with this trend, the concept of optimizing real number arithmetic has been raised. In this paper, we evaluate the performance of the noble number system named posit on neural networks by analyzing the execution of approximate exponential functions, which is fundamental to several activation functions, with posit32 and float32. To implement the functions with posit arithmetic, we designed the software posit library consisting of basic arithmetic operations and conversion operations from/to C standard data types. The result shows that posit arithmetic reduces the average relative error rate by up to 87.12% on the exponential function.

AB - Recent advances in semiconductor technology lead to ongoing applications to adopt complex techniques based on neural networks. In line with this trend, the concept of optimizing real number arithmetic has been raised. In this paper, we evaluate the performance of the noble number system named posit on neural networks by analyzing the execution of approximate exponential functions, which is fundamental to several activation functions, with posit32 and float32. To implement the functions with posit arithmetic, we designed the software posit library consisting of basic arithmetic operations and conversion operations from/to C standard data types. The result shows that posit arithmetic reduces the average relative error rate by up to 87.12% on the exponential function.

KW - activation functions

KW - exponential approximation

KW - IEEE-754

KW - posit

KW - real number arithmetic

UR - http://www.scopus.com/inward/record.url?scp=85148418519&partnerID=8YFLogxK

U2 - 10.1109/ISOCC56007.2022.10031524

DO - 10.1109/ISOCC56007.2022.10031524

M3 - Conference contribution

AN - SCOPUS:85148418519

T3 - Proceedings - International SoC Design Conference 2022, ISOCC 2022

SP - 358

EP - 359

BT - Proceedings - International SoC Design Conference 2022, ISOCC 2022

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 19th International System-on-Chip Design Conference, ISOCC 2022

Y2 - 19 October 2022 through 22 October 2022

ER -

Oh HW, Jeong WS, Lee SE. Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions. In Proceedings - International SoC Design Conference 2022, ISOCC 2022. Institute of Electrical and Electronics Engineers Inc. 2022. p. 358-359. (Proceedings - International SoC Design Conference 2022, ISOCC 2022). doi: 10.1109/ISOCC56007.2022.10031524

Evaluation of Posit Arithmetic on Machine Learning based on Approximate Exponential Functions

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this