A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array

Mingu Kang, Sujan K. Gonugondla, Naresh R Shanbhag

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

Abstract

This paper presents IC realization of a random forest (RF) machine learning classifier. Algorithm-architecture-circuit is co-optimized to minimize the energy-delay product (EDP). Deterministic subsampling (DSS) and balanced decision trees result in reduced interconnect complexity and avoid irregular memory accesses. Low-swing analog in-memory computations embedded in a standard 6T SRAM enable massively parallel processing thereby minimizing the memory fetches and reducing the EDP further. The 65nm CMOS prototype achieves a 6.8× lower EDP compared to a conventional design at the same accuracy (94%) for an 8-class traffic sign recognition problem.

Original languageEnglish (US)
Title of host publicationESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages263-266
Number of pages4
ISBN (Electronic)9781509050253
DOIs
StatePublished - Nov 2 2017
Event43rd IEEE European Solid State Circuits Conference, ESSCIRC 2017 - Leuven, Belgium
Duration: Sep 11 2017Sep 14 2017

Publication series

NameESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference

Other

Other43rd IEEE European Solid State Circuits Conference, ESSCIRC 2017
CountryBelgium
CityLeuven
Period9/11/179/14/17

Fingerprint

Static random access storage
classifiers
Classifiers
Data storage equipment
products
Traffic signs
machine learning
Decision trees
traffic
Learning systems
energy
CMOS
prototypes
analogs
Networks (circuits)
Processing

Keywords

  • In-memory computing
  • Machine learning
  • Pattern recognition
  • Random forest
  • Traffic sign recognition

ASJC Scopus subject areas

  • Instrumentation
  • Hardware and Architecture
  • Electrical and Electronic Engineering
  • Computer Networks and Communications

Cite this

Kang, M., Gonugondla, S. K., & Shanbhag, N. R. (2017). A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array. In ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference (pp. 263-266). [8094576] (ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ESSCIRC.2017.8094576

A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array. / Kang, Mingu; Gonugondla, Sujan K.; Shanbhag, Naresh R.

ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference. Institute of Electrical and Electronics Engineers Inc., 2017. p. 263-266 8094576 (ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference).

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

Kang, M, Gonugondla, SK & Shanbhag, NR 2017, A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array. in ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference., 8094576, ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference, Institute of Electrical and Electronics Engineers Inc., pp. 263-266, 43rd IEEE European Solid State Circuits Conference, ESSCIRC 2017, Leuven, Belgium, 9/11/17. https://doi.org/10.1109/ESSCIRC.2017.8094576
Kang M, Gonugondla SK, Shanbhag NR. A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array. In ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference. Institute of Electrical and Electronics Engineers Inc. 2017. p. 263-266. 8094576. (ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference). https://doi.org/10.1109/ESSCIRC.2017.8094576
Kang, Mingu ; Gonugondla, Sujan K. ; Shanbhag, Naresh R. / A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array. ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 263-266 (ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference).
@inproceedings{ab493a90c7554e269401a7cec3aa45d8,
title = "A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array",
abstract = "This paper presents IC realization of a random forest (RF) machine learning classifier. Algorithm-architecture-circuit is co-optimized to minimize the energy-delay product (EDP). Deterministic subsampling (DSS) and balanced decision trees result in reduced interconnect complexity and avoid irregular memory accesses. Low-swing analog in-memory computations embedded in a standard 6T SRAM enable massively parallel processing thereby minimizing the memory fetches and reducing the EDP further. The 65nm CMOS prototype achieves a 6.8× lower EDP compared to a conventional design at the same accuracy (94{\%}) for an 8-class traffic sign recognition problem.",
keywords = "In-memory computing, Machine learning, Pattern recognition, Random forest, Traffic sign recognition",
author = "Mingu Kang and Gonugondla, {Sujan K.} and Shanbhag, {Naresh R}",
year = "2017",
month = "11",
day = "2",
doi = "10.1109/ESSCIRC.2017.8094576",
language = "English (US)",
series = "ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "263--266",
booktitle = "ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference",
address = "United States",

}

TY - GEN

T1 - A 19.4 nJ/decision 364K decisions/s in-memory random forest classifier in 6T SRAM array

AU - Kang, Mingu

AU - Gonugondla, Sujan K.

AU - Shanbhag, Naresh R

PY - 2017/11/2

Y1 - 2017/11/2

N2 - This paper presents IC realization of a random forest (RF) machine learning classifier. Algorithm-architecture-circuit is co-optimized to minimize the energy-delay product (EDP). Deterministic subsampling (DSS) and balanced decision trees result in reduced interconnect complexity and avoid irregular memory accesses. Low-swing analog in-memory computations embedded in a standard 6T SRAM enable massively parallel processing thereby minimizing the memory fetches and reducing the EDP further. The 65nm CMOS prototype achieves a 6.8× lower EDP compared to a conventional design at the same accuracy (94%) for an 8-class traffic sign recognition problem.

AB - This paper presents IC realization of a random forest (RF) machine learning classifier. Algorithm-architecture-circuit is co-optimized to minimize the energy-delay product (EDP). Deterministic subsampling (DSS) and balanced decision trees result in reduced interconnect complexity and avoid irregular memory accesses. Low-swing analog in-memory computations embedded in a standard 6T SRAM enable massively parallel processing thereby minimizing the memory fetches and reducing the EDP further. The 65nm CMOS prototype achieves a 6.8× lower EDP compared to a conventional design at the same accuracy (94%) for an 8-class traffic sign recognition problem.

KW - In-memory computing

KW - Machine learning

KW - Pattern recognition

KW - Random forest

KW - Traffic sign recognition

UR - http://www.scopus.com/inward/record.url?scp=85040559508&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040559508&partnerID=8YFLogxK

U2 - 10.1109/ESSCIRC.2017.8094576

DO - 10.1109/ESSCIRC.2017.8094576

M3 - Conference contribution

AN - SCOPUS:85040559508

T3 - ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference

SP - 263

EP - 266

BT - ESSCIRC 2017 - 43rd IEEE European Solid State Circuits Conference

PB - Institute of Electrical and Electronics Engineers Inc.

ER -