Digital CMOS neuromorphic processor design featuring unsupervised online learning

Jae-sun Seo, Mingoo Seok

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

8 Citations (Scopus)

Abstract

The compute-intensive and power-efficient brain has been a source of inspiration for a broad range of neural networks to solve recognition and classification tasks. Compared to the supervised deep neural networks (DNNs) that have been very successful on well-defined labeled datasets, bio-plausible spiking neural networks (SNNs) with unsupervised learning rules could be well-suited for training and learning representations from the massive amount of unlabeled data. To design dense and low-power hardware for such unsupervised SNNs, we employ digital CMOS circuits for neuromorphic processors, which can exploit transistor scaling and dynamic voltage scaling to the utmost. As exemplary works, we present two neuromorphic processor designs. First, a 45nm neuromorphic chip is designed for a small-scale network of spiking neurons. Through tight integration of memory (64k SRAM synapses) and computation (256 digital neurons), the chip demonstrates on-chip learning on pattern recognition tasks down to 0.53V supply. Secondly, a 65nm neuromorphic processor that performs unsupervised on-line spike-clustering for brain sensing applications is implemented with 1.2k digital neurons and 4.7k latch-based synapses. The processor exhibits a power consumption of 9.3μW/ch at 0.3V supply. Synapse hardware precision, efficient synapse memory array access, overfitting, and voltage scaling will be discussed for dense and power-efficient on-chip learning for CMOS spiking neural networks.

Original languageEnglish (US)
Title of host publicationIEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC
PublisherIEEE Computer Society
Pages49-51
Number of pages3
Volume2015-October
ISBN (Print)9781467391405
DOIs
StatePublished - Oct 30 2015
Event23rd IFIP/IEEE International Conference on Very Large Scale Integration, VLSI-SoC 2015 - Daejeon, Korea, Republic of
Duration: Oct 5 2015Oct 7 2015

Other

Other23rd IFIP/IEEE International Conference on Very Large Scale Integration, VLSI-SoC 2015
CountryKorea, Republic of
CityDaejeon
Period10/5/1510/7/15

Fingerprint

Neurons
Neural networks
Brain
Data storage equipment
Unsupervised learning
Static random access storage
Complex networks
Computer hardware
Pattern recognition
Transistors
Electric power utilization
Hardware
Networks (circuits)
Voltage scaling
Deep neural networks

Keywords

  • CMOS
  • digital circuits
  • low-power
  • low-voltage
  • neuromorphic computing
  • on-chip learning
  • spiking neural networks
  • unsupervised learning

ASJC Scopus subject areas

  • Hardware and Architecture
  • Software
  • Electrical and Electronic Engineering

Cite this

Seo, J., & Seok, M. (2015). Digital CMOS neuromorphic processor design featuring unsupervised online learning. In IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC (Vol. 2015-October, pp. 49-51). [7314390] IEEE Computer Society. https://doi.org/10.1109/VLSI-SoC.2015.7314390

Digital CMOS neuromorphic processor design featuring unsupervised online learning. / Seo, Jae-sun; Seok, Mingoo.

IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC. Vol. 2015-October IEEE Computer Society, 2015. p. 49-51 7314390.

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

Seo, J & Seok, M 2015, Digital CMOS neuromorphic processor design featuring unsupervised online learning. in IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC. vol. 2015-October, 7314390, IEEE Computer Society, pp. 49-51, 23rd IFIP/IEEE International Conference on Very Large Scale Integration, VLSI-SoC 2015, Daejeon, Korea, Republic of, 10/5/15. https://doi.org/10.1109/VLSI-SoC.2015.7314390
Seo J, Seok M. Digital CMOS neuromorphic processor design featuring unsupervised online learning. In IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC. Vol. 2015-October. IEEE Computer Society. 2015. p. 49-51. 7314390 https://doi.org/10.1109/VLSI-SoC.2015.7314390
Seo, Jae-sun ; Seok, Mingoo. / Digital CMOS neuromorphic processor design featuring unsupervised online learning. IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC. Vol. 2015-October IEEE Computer Society, 2015. pp. 49-51
@inproceedings{c85d9ea1eb4c49a0ac9b063c1db94d18,
title = "Digital CMOS neuromorphic processor design featuring unsupervised online learning",
abstract = "The compute-intensive and power-efficient brain has been a source of inspiration for a broad range of neural networks to solve recognition and classification tasks. Compared to the supervised deep neural networks (DNNs) that have been very successful on well-defined labeled datasets, bio-plausible spiking neural networks (SNNs) with unsupervised learning rules could be well-suited for training and learning representations from the massive amount of unlabeled data. To design dense and low-power hardware for such unsupervised SNNs, we employ digital CMOS circuits for neuromorphic processors, which can exploit transistor scaling and dynamic voltage scaling to the utmost. As exemplary works, we present two neuromorphic processor designs. First, a 45nm neuromorphic chip is designed for a small-scale network of spiking neurons. Through tight integration of memory (64k SRAM synapses) and computation (256 digital neurons), the chip demonstrates on-chip learning on pattern recognition tasks down to 0.53V supply. Secondly, a 65nm neuromorphic processor that performs unsupervised on-line spike-clustering for brain sensing applications is implemented with 1.2k digital neurons and 4.7k latch-based synapses. The processor exhibits a power consumption of 9.3μW/ch at 0.3V supply. Synapse hardware precision, efficient synapse memory array access, overfitting, and voltage scaling will be discussed for dense and power-efficient on-chip learning for CMOS spiking neural networks.",
keywords = "CMOS, digital circuits, low-power, low-voltage, neuromorphic computing, on-chip learning, spiking neural networks, unsupervised learning",
author = "Jae-sun Seo and Mingoo Seok",
year = "2015",
month = "10",
day = "30",
doi = "10.1109/VLSI-SoC.2015.7314390",
language = "English (US)",
isbn = "9781467391405",
volume = "2015-October",
pages = "49--51",
booktitle = "IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Digital CMOS neuromorphic processor design featuring unsupervised online learning

AU - Seo, Jae-sun

AU - Seok, Mingoo

PY - 2015/10/30

Y1 - 2015/10/30

N2 - The compute-intensive and power-efficient brain has been a source of inspiration for a broad range of neural networks to solve recognition and classification tasks. Compared to the supervised deep neural networks (DNNs) that have been very successful on well-defined labeled datasets, bio-plausible spiking neural networks (SNNs) with unsupervised learning rules could be well-suited for training and learning representations from the massive amount of unlabeled data. To design dense and low-power hardware for such unsupervised SNNs, we employ digital CMOS circuits for neuromorphic processors, which can exploit transistor scaling and dynamic voltage scaling to the utmost. As exemplary works, we present two neuromorphic processor designs. First, a 45nm neuromorphic chip is designed for a small-scale network of spiking neurons. Through tight integration of memory (64k SRAM synapses) and computation (256 digital neurons), the chip demonstrates on-chip learning on pattern recognition tasks down to 0.53V supply. Secondly, a 65nm neuromorphic processor that performs unsupervised on-line spike-clustering for brain sensing applications is implemented with 1.2k digital neurons and 4.7k latch-based synapses. The processor exhibits a power consumption of 9.3μW/ch at 0.3V supply. Synapse hardware precision, efficient synapse memory array access, overfitting, and voltage scaling will be discussed for dense and power-efficient on-chip learning for CMOS spiking neural networks.

AB - The compute-intensive and power-efficient brain has been a source of inspiration for a broad range of neural networks to solve recognition and classification tasks. Compared to the supervised deep neural networks (DNNs) that have been very successful on well-defined labeled datasets, bio-plausible spiking neural networks (SNNs) with unsupervised learning rules could be well-suited for training and learning representations from the massive amount of unlabeled data. To design dense and low-power hardware for such unsupervised SNNs, we employ digital CMOS circuits for neuromorphic processors, which can exploit transistor scaling and dynamic voltage scaling to the utmost. As exemplary works, we present two neuromorphic processor designs. First, a 45nm neuromorphic chip is designed for a small-scale network of spiking neurons. Through tight integration of memory (64k SRAM synapses) and computation (256 digital neurons), the chip demonstrates on-chip learning on pattern recognition tasks down to 0.53V supply. Secondly, a 65nm neuromorphic processor that performs unsupervised on-line spike-clustering for brain sensing applications is implemented with 1.2k digital neurons and 4.7k latch-based synapses. The processor exhibits a power consumption of 9.3μW/ch at 0.3V supply. Synapse hardware precision, efficient synapse memory array access, overfitting, and voltage scaling will be discussed for dense and power-efficient on-chip learning for CMOS spiking neural networks.

KW - CMOS

KW - digital circuits

KW - low-power

KW - low-voltage

KW - neuromorphic computing

KW - on-chip learning

KW - spiking neural networks

KW - unsupervised learning

UR - http://www.scopus.com/inward/record.url?scp=84960099852&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84960099852&partnerID=8YFLogxK

U2 - 10.1109/VLSI-SoC.2015.7314390

DO - 10.1109/VLSI-SoC.2015.7314390

M3 - Conference contribution

AN - SCOPUS:84960099852

SN - 9781467391405

VL - 2015-October

SP - 49

EP - 51

BT - IEEE/IFIP International Conference on VLSI and System-on-Chip, VLSI-SoC

PB - IEEE Computer Society

ER -