Special Session: Does Approximation Make Testing Harder (or Easier)?

R. Iris Bahar, Ulya Karpuzcu, Sasa Misailovic

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many important application domains, including machine learning, feature intrinsically noise tolerant algorithms. These algorithms process massive, yet noisy and redundant data, by probabilistic and often iterative techniques. As a result, there is a range of valid outputs rather than a single golden value. While this may translate into relaxed constraints for testing and verification of approximate systems, distinguishing actual design bugs from what is being approximated also becomes harder. In this paper, using representative case studies, we pose several challenges for the test and verification community as approximate computing becomes more prevalent as a design of choice in order to achieve performance gains, power or energy savings, improved reliability or reduced software and/or hardware complexity.

Original languageEnglish (US)
Title of host publication2019 IEEE 37th VLSI Test Symposium, VTS 2019
PublisherIEEE Computer Society
ISBN (Electronic)9781728111704
DOIs
StatePublished - Apr 2019
Event37th IEEE VLSI Test Symposium, VTS 2019 - Monterey, United States
Duration: Apr 23 2019Apr 25 2019

Publication series

NameProceedings of the IEEE VLSI Test Symposium
Volume2019-April

Conference

Conference37th IEEE VLSI Test Symposium, VTS 2019
CountryUnited States
CityMonterey
Period4/23/194/25/19

Fingerprint

Testing
Learning systems
Energy conservation
Hardware

ASJC Scopus subject areas

  • Computer Science Applications
  • Electrical and Electronic Engineering

Cite this

Bahar, R. I., Karpuzcu, U., & Misailovic, S. (2019). Special Session: Does Approximation Make Testing Harder (or Easier)? In 2019 IEEE 37th VLSI Test Symposium, VTS 2019 [8758649] (Proceedings of the IEEE VLSI Test Symposium; Vol. 2019-April). IEEE Computer Society. https://doi.org/10.1109/VTS.2019.8758649

Special Session : Does Approximation Make Testing Harder (or Easier)? / Bahar, R. Iris; Karpuzcu, Ulya; Misailovic, Sasa.

2019 IEEE 37th VLSI Test Symposium, VTS 2019. IEEE Computer Society, 2019. 8758649 (Proceedings of the IEEE VLSI Test Symposium; Vol. 2019-April).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bahar, RI, Karpuzcu, U & Misailovic, S 2019, Special Session: Does Approximation Make Testing Harder (or Easier)? in 2019 IEEE 37th VLSI Test Symposium, VTS 2019., 8758649, Proceedings of the IEEE VLSI Test Symposium, vol. 2019-April, IEEE Computer Society, 37th IEEE VLSI Test Symposium, VTS 2019, Monterey, United States, 4/23/19. https://doi.org/10.1109/VTS.2019.8758649
Bahar RI, Karpuzcu U, Misailovic S. Special Session: Does Approximation Make Testing Harder (or Easier)? In 2019 IEEE 37th VLSI Test Symposium, VTS 2019. IEEE Computer Society. 2019. 8758649. (Proceedings of the IEEE VLSI Test Symposium). https://doi.org/10.1109/VTS.2019.8758649
Bahar, R. Iris ; Karpuzcu, Ulya ; Misailovic, Sasa. / Special Session : Does Approximation Make Testing Harder (or Easier)?. 2019 IEEE 37th VLSI Test Symposium, VTS 2019. IEEE Computer Society, 2019. (Proceedings of the IEEE VLSI Test Symposium).
@inproceedings{dddb195225204f4e90ed303318c6f5cd,
title = "Special Session: Does Approximation Make Testing Harder (or Easier)?",
abstract = "Many important application domains, including machine learning, feature intrinsically noise tolerant algorithms. These algorithms process massive, yet noisy and redundant data, by probabilistic and often iterative techniques. As a result, there is a range of valid outputs rather than a single golden value. While this may translate into relaxed constraints for testing and verification of approximate systems, distinguishing actual design bugs from what is being approximated also becomes harder. In this paper, using representative case studies, we pose several challenges for the test and verification community as approximate computing becomes more prevalent as a design of choice in order to achieve performance gains, power or energy savings, improved reliability or reduced software and/or hardware complexity.",
author = "Bahar, {R. Iris} and Ulya Karpuzcu and Sasa Misailovic",
year = "2019",
month = "4",
doi = "10.1109/VTS.2019.8758649",
language = "English (US)",
series = "Proceedings of the IEEE VLSI Test Symposium",
publisher = "IEEE Computer Society",
booktitle = "2019 IEEE 37th VLSI Test Symposium, VTS 2019",

}

TY - GEN

T1 - Special Session

T2 - Does Approximation Make Testing Harder (or Easier)?

AU - Bahar, R. Iris

AU - Karpuzcu, Ulya

AU - Misailovic, Sasa

PY - 2019/4

Y1 - 2019/4

N2 - Many important application domains, including machine learning, feature intrinsically noise tolerant algorithms. These algorithms process massive, yet noisy and redundant data, by probabilistic and often iterative techniques. As a result, there is a range of valid outputs rather than a single golden value. While this may translate into relaxed constraints for testing and verification of approximate systems, distinguishing actual design bugs from what is being approximated also becomes harder. In this paper, using representative case studies, we pose several challenges for the test and verification community as approximate computing becomes more prevalent as a design of choice in order to achieve performance gains, power or energy savings, improved reliability or reduced software and/or hardware complexity.

AB - Many important application domains, including machine learning, feature intrinsically noise tolerant algorithms. These algorithms process massive, yet noisy and redundant data, by probabilistic and often iterative techniques. As a result, there is a range of valid outputs rather than a single golden value. While this may translate into relaxed constraints for testing and verification of approximate systems, distinguishing actual design bugs from what is being approximated also becomes harder. In this paper, using representative case studies, we pose several challenges for the test and verification community as approximate computing becomes more prevalent as a design of choice in order to achieve performance gains, power or energy savings, improved reliability or reduced software and/or hardware complexity.

UR - http://www.scopus.com/inward/record.url?scp=85069172442&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85069172442&partnerID=8YFLogxK

U2 - 10.1109/VTS.2019.8758649

DO - 10.1109/VTS.2019.8758649

M3 - Conference contribution

AN - SCOPUS:85069172442

T3 - Proceedings of the IEEE VLSI Test Symposium

BT - 2019 IEEE 37th VLSI Test Symposium, VTS 2019

PB - IEEE Computer Society

ER -