Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization

Xianbo Yang; Hakki M. Torun; Junyan Tang; Pavel Roy Paladhi; Yanyan Zhang; Wiren D. Becker; Jose A. Hejase; Madhavan Swaminathan

doi:10.1109/EPEPS51341.2021.9609205

Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization

Xianbo Yang, Hakki M. Torun, Junyan Tang, Pavel Roy Paladhi, Yanyan Zhang, Wiren D. Becker, Jose A. Hejase, Madhavan Swaminathan

Electrical Engineering

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

1 Scopus citations

Abstract

This work realizes the parallelization of Bayesian Active Learning using Dropout (BAL-DO) and is successfully applied for optimizing equalization settings for high-speed channel receivers (RX). In this paper, parallel BAL-DO can achieve the largest horizontal eye (HEYE) opening and its corresponding equalization setting 12 times faster, on average, than the previously reported sequential BAL-DO. This corresponds to an average of 40-50 times faster than standard exhaustive time-domain simulations. Moreover, the HEYE prediction accuracy across the whole design space is close to 3 times better while using parallelization than sequential BAL-DO. With these outstanding results, parallel BAL-DO dramatically improves the efficiency for RX equalization optimization and greatly reduces the computing resources and time needed.

Original language	English (US)
Title of host publication	EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9781665442695
DOIs	https://doi.org/10.1109/EPEPS51341.2021.9609205
State	Published - 2021
Event	30th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2021 - Austin, United States Duration: Oct 17 2021 → Oct 20 2021

Publication series

Name	EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems

Conference

Conference	30th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2021
Country/Territory	United States
City	Austin
Period	10/17/21 → 10/20/21

All Science Journal Classification (ASJC) codes

Safety, Risk, Reliability and Quality
Electronic, Optical and Magnetic Materials
Hardware and Architecture
Electrical and Electronic Engineering

Access to Document

10.1109/EPEPS51341.2021.9609205

Cite this

Yang, X., Torun, H. M., Tang, J., Paladhi, P. R., Zhang, Y., Becker, W. D., Hejase, J. A., & Swaminathan, M. (2021). Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization. In EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems (EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/EPEPS51341.2021.9609205

Yang, Xianbo ; Torun, Hakki M. ; Tang, Junyan et al. / Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization. EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems. Institute of Electrical and Electronics Engineers Inc., 2021. (EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems).

@inproceedings{4809d11fa24f4a31a077cad324dbed07,

title = "Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization",

abstract = "This work realizes the parallelization of Bayesian Active Learning using Dropout (BAL-DO) and is successfully applied for optimizing equalization settings for high-speed channel receivers (RX). In this paper, parallel BAL-DO can achieve the largest horizontal eye (HEYE) opening and its corresponding equalization setting 12 times faster, on average, than the previously reported sequential BAL-DO. This corresponds to an average of 40-50 times faster than standard exhaustive time-domain simulations. Moreover, the HEYE prediction accuracy across the whole design space is close to 3 times better while using parallelization than sequential BAL-DO. With these outstanding results, parallel BAL-DO dramatically improves the efficiency for RX equalization optimization and greatly reduces the computing resources and time needed.",

author = "Xianbo Yang and Torun, {Hakki M.} and Junyan Tang and Paladhi, {Pavel Roy} and Yanyan Zhang and Becker, {Wiren D.} and Hejase, {Jose A.} and Madhavan Swaminathan",

note = "Publisher Copyright: {\textcopyright} 2021 IEEE.; 30th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2021 ; Conference date: 17-10-2021 Through 20-10-2021",

year = "2021",

doi = "10.1109/EPEPS51341.2021.9609205",

language = "English (US)",

series = "EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems",

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

booktitle = "EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems",

address = "United States",

}

Yang, X, Torun, HM, Tang, J, Paladhi, PR, Zhang, Y, Becker, WD, Hejase, JA & Swaminathan, M 2021, Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization. in EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems. EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems, Institute of Electrical and Electronics Engineers Inc., 30th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2021, Austin, United States, 10/17/21. https://doi.org/10.1109/EPEPS51341.2021.9609205

Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization. / Yang, Xianbo; Torun, Hakki M.; Tang, Junyan et al.
EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems. Institute of Electrical and Electronics Engineers Inc., 2021. (EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems).

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

TY - GEN

T1 - Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization

AU - Yang, Xianbo

AU - Torun, Hakki M.

AU - Tang, Junyan

AU - Paladhi, Pavel Roy

AU - Zhang, Yanyan

AU - Becker, Wiren D.

AU - Hejase, Jose A.

AU - Swaminathan, Madhavan

PY - 2021

Y1 - 2021

N2 - This work realizes the parallelization of Bayesian Active Learning using Dropout (BAL-DO) and is successfully applied for optimizing equalization settings for high-speed channel receivers (RX). In this paper, parallel BAL-DO can achieve the largest horizontal eye (HEYE) opening and its corresponding equalization setting 12 times faster, on average, than the previously reported sequential BAL-DO. This corresponds to an average of 40-50 times faster than standard exhaustive time-domain simulations. Moreover, the HEYE prediction accuracy across the whole design space is close to 3 times better while using parallelization than sequential BAL-DO. With these outstanding results, parallel BAL-DO dramatically improves the efficiency for RX equalization optimization and greatly reduces the computing resources and time needed.

AB - This work realizes the parallelization of Bayesian Active Learning using Dropout (BAL-DO) and is successfully applied for optimizing equalization settings for high-speed channel receivers (RX). In this paper, parallel BAL-DO can achieve the largest horizontal eye (HEYE) opening and its corresponding equalization setting 12 times faster, on average, than the previously reported sequential BAL-DO. This corresponds to an average of 40-50 times faster than standard exhaustive time-domain simulations. Moreover, the HEYE prediction accuracy across the whole design space is close to 3 times better while using parallelization than sequential BAL-DO. With these outstanding results, parallel BAL-DO dramatically improves the efficiency for RX equalization optimization and greatly reduces the computing resources and time needed.

UR - http://www.scopus.com/inward/record.url?scp=85123195929&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85123195929&partnerID=8YFLogxK

U2 - 10.1109/EPEPS51341.2021.9609205

DO - 10.1109/EPEPS51341.2021.9609205

M3 - Conference contribution

AN - SCOPUS:85123195929

T3 - EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems

BT - EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 30th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2021

Y2 - 17 October 2021 through 20 October 2021

ER -

Yang X, Torun HM, Tang J, Paladhi PR, Zhang Y, Becker WD et al. Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization. In EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems. Institute of Electrical and Electronics Engineers Inc. 2021. (EPEPS 2021 - IEEE 30th Conference on Electrical Performance of Electronic Packaging and Systems). doi: 10.1109/EPEPS51341.2021.9609205

Parallel Bayesian Active Learning using Dropout for Optimizing High-Speed Channel Equalization

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this