EPOC: A 28-nm 5.3 pJ/SOP Event-driven Parallel Neuromorphic Hardware with Neuromodulation-based Online Learning

Bio-inspired neuromorphic hardware with learning ability is highly promising to achieve human-like intelligence, particularly in terms of high energy efficiency and strong environmental adaptability. Though many customized prototypes have demonstrated learning ability, learning on neuromorphic hardw...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on biomedical circuits and systems 2024-10, Vol.PP, p.1-16
Hauptverfasser: Chen, Faquan, Tian, Qingyang, Xie, Lisheng, Zhou, Yifan, Wu, Ziren, Wu, Liangshun, Ying, Rendong, Wen, Fei, Liu, Peilin
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 16
container_issue
container_start_page 1
container_title IEEE transactions on biomedical circuits and systems
container_volume PP
creator Chen, Faquan
Tian, Qingyang
Xie, Lisheng
Zhou, Yifan
Wu, Ziren
Wu, Liangshun
Ying, Rendong
Wen, Fei
Liu, Peilin
description Bio-inspired neuromorphic hardware with learning ability is highly promising to achieve human-like intelligence, particularly in terms of high energy efficiency and strong environmental adaptability. Though many customized prototypes have demonstrated learning ability, learning on neuromorphic hardware still lacks a bio-plausible and unified learning framework, and inherent spike-based sparsity and parallelism have not been fully exploited, which fundamentally limits their computational efficiency and scale. Therefore, we develop a unified, event-driven, and massively parallel multi-core neuromorphic online learning processor, namely EPOC. We present a neuromodulation-based neuromorphic online learning framework to unify various learning algorithms, and EPOC supports high-accuracy local/global supervised Spike Neural Network (SNN) learning with a low-memory-demand streaming single-sample learning strategy through different neuromodulator formulations. EPOC leverages a novel event-driven computation method that fully exploits spike-based sparsity throughout the forward-backward learning phases, and parallel multi-channel and multi-core computing architecture, bringing 9.9× time efficiency improvement compared with the baseline architecture. We synthesize EPOC in a 28-nm CMOS process and perform extensive benchmarking. EPOC achieves state-of-the-art learning accuracy of 99.2%, 98.2%, and 94.3% on the MNIST, NMNIST, and DVS-Gesture benchmarks, respectively. Local-learning EPOC achieves 2.9× time efficiency improvement compared with the global learning counterpart. EPOC operates at a typical clock frequency of 100 MHz, providing a peak 328 GOPS/51 GSOPS throughput and a 5.3 pJ/SOP energy efficiency.
doi_str_mv 10.1109/TBCAS.2024.3470520
format Article
fullrecord <record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_ieee_primary_10704604</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10704604</ieee_id><sourcerecordid>3112525794</sourcerecordid><originalsourceid>FETCH-LOGICAL-c1204-de3e36378d5d674528c499b5bbf3cc859eb53fa21462dae6e5c5a84b0cadbe8e3</originalsourceid><addsrcrecordid>eNpNkEtPAjEUhRujEUT_gDGmSzcDfc5M3SFB0RAhAdeTTnuRmnlgy0j894KAcXVucr9zFh9C15R0KSWqN38Y9GddRpjocpEQycgJalMlSKSUIqe7m7NISCFb6CKED0JkzBQ7Ry2uuIylEm1UD6eTwT3uY5ZGVYlll-PVS282meLhF1TryHq3TTzVXhcFFPgVGl-XtV8tncEj7e1Ge8Abt14eX7Yp9NrVVZTrABZPqsJVgMegfeWq90t0ttBFgKtDdtDb43A-GEXjydPzoD-ODGVERBY48JgnqZU2ToRkqRFK5TLPF9yYVCrIJV9oRkXMrIYYpJE6FTkx2uaQAu-gu_3uytefDYR1VrpgoCh0BXUTMk4pk0wmSmxRtkeNr0PwsMhW3pXaf2eUZDvR2a_obCc6O4jelm4P-01egv2rHM1ugZs94ADg32JCREwE_wHzvIF3</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3112525794</pqid></control><display><type>article</type><title>EPOC: A 28-nm 5.3 pJ/SOP Event-driven Parallel Neuromorphic Hardware with Neuromodulation-based Online Learning</title><source>IEEE/IET Electronic Library</source><creator>Chen, Faquan ; Tian, Qingyang ; Xie, Lisheng ; Zhou, Yifan ; Wu, Ziren ; Wu, Liangshun ; Ying, Rendong ; Wen, Fei ; Liu, Peilin</creator><creatorcontrib>Chen, Faquan ; Tian, Qingyang ; Xie, Lisheng ; Zhou, Yifan ; Wu, Ziren ; Wu, Liangshun ; Ying, Rendong ; Wen, Fei ; Liu, Peilin</creatorcontrib><description>Bio-inspired neuromorphic hardware with learning ability is highly promising to achieve human-like intelligence, particularly in terms of high energy efficiency and strong environmental adaptability. Though many customized prototypes have demonstrated learning ability, learning on neuromorphic hardware still lacks a bio-plausible and unified learning framework, and inherent spike-based sparsity and parallelism have not been fully exploited, which fundamentally limits their computational efficiency and scale. Therefore, we develop a unified, event-driven, and massively parallel multi-core neuromorphic online learning processor, namely EPOC. We present a neuromodulation-based neuromorphic online learning framework to unify various learning algorithms, and EPOC supports high-accuracy local/global supervised Spike Neural Network (SNN) learning with a low-memory-demand streaming single-sample learning strategy through different neuromodulator formulations. EPOC leverages a novel event-driven computation method that fully exploits spike-based sparsity throughout the forward-backward learning phases, and parallel multi-channel and multi-core computing architecture, bringing 9.9× time efficiency improvement compared with the baseline architecture. We synthesize EPOC in a 28-nm CMOS process and perform extensive benchmarking. EPOC achieves state-of-the-art learning accuracy of 99.2%, 98.2%, and 94.3% on the MNIST, NMNIST, and DVS-Gesture benchmarks, respectively. Local-learning EPOC achieves 2.9× time efficiency improvement compared with the global learning counterpart. EPOC operates at a typical clock frequency of 100 MHz, providing a peak 328 GOPS/51 GSOPS throughput and a 5.3 pJ/SOP energy efficiency.</description><identifier>ISSN: 1932-4545</identifier><identifier>ISSN: 1940-9990</identifier><identifier>EISSN: 1940-9990</identifier><identifier>DOI: 10.1109/TBCAS.2024.3470520</identifier><identifier>PMID: 39356594</identifier><identifier>CODEN: ITBCCW</identifier><language>eng</language><publisher>United States: IEEE</publisher><subject>artificial neural network ; Biological neural networks ; Complexity theory ; Computational efficiency ; Hardware ; Memory management ; neuromodulation ; Neuromorphic hardware ; Neuromorphics ; Neurons ; on-chip learning ; Parallel processing ; parallelism ; spike neural network ; spike-based sparsity ; System-on-chip ; Training</subject><ispartof>IEEE transactions on biomedical circuits and systems, 2024-10, Vol.PP, p.1-16</ispartof><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10704604$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,776,780,792,27901,27902,54733</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10704604$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/39356594$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Chen, Faquan</creatorcontrib><creatorcontrib>Tian, Qingyang</creatorcontrib><creatorcontrib>Xie, Lisheng</creatorcontrib><creatorcontrib>Zhou, Yifan</creatorcontrib><creatorcontrib>Wu, Ziren</creatorcontrib><creatorcontrib>Wu, Liangshun</creatorcontrib><creatorcontrib>Ying, Rendong</creatorcontrib><creatorcontrib>Wen, Fei</creatorcontrib><creatorcontrib>Liu, Peilin</creatorcontrib><title>EPOC: A 28-nm 5.3 pJ/SOP Event-driven Parallel Neuromorphic Hardware with Neuromodulation-based Online Learning</title><title>IEEE transactions on biomedical circuits and systems</title><addtitle>TBCAS</addtitle><addtitle>IEEE Trans Biomed Circuits Syst</addtitle><description>Bio-inspired neuromorphic hardware with learning ability is highly promising to achieve human-like intelligence, particularly in terms of high energy efficiency and strong environmental adaptability. Though many customized prototypes have demonstrated learning ability, learning on neuromorphic hardware still lacks a bio-plausible and unified learning framework, and inherent spike-based sparsity and parallelism have not been fully exploited, which fundamentally limits their computational efficiency and scale. Therefore, we develop a unified, event-driven, and massively parallel multi-core neuromorphic online learning processor, namely EPOC. We present a neuromodulation-based neuromorphic online learning framework to unify various learning algorithms, and EPOC supports high-accuracy local/global supervised Spike Neural Network (SNN) learning with a low-memory-demand streaming single-sample learning strategy through different neuromodulator formulations. EPOC leverages a novel event-driven computation method that fully exploits spike-based sparsity throughout the forward-backward learning phases, and parallel multi-channel and multi-core computing architecture, bringing 9.9× time efficiency improvement compared with the baseline architecture. We synthesize EPOC in a 28-nm CMOS process and perform extensive benchmarking. EPOC achieves state-of-the-art learning accuracy of 99.2%, 98.2%, and 94.3% on the MNIST, NMNIST, and DVS-Gesture benchmarks, respectively. Local-learning EPOC achieves 2.9× time efficiency improvement compared with the global learning counterpart. EPOC operates at a typical clock frequency of 100 MHz, providing a peak 328 GOPS/51 GSOPS throughput and a 5.3 pJ/SOP energy efficiency.</description><subject>artificial neural network</subject><subject>Biological neural networks</subject><subject>Complexity theory</subject><subject>Computational efficiency</subject><subject>Hardware</subject><subject>Memory management</subject><subject>neuromodulation</subject><subject>Neuromorphic hardware</subject><subject>Neuromorphics</subject><subject>Neurons</subject><subject>on-chip learning</subject><subject>Parallel processing</subject><subject>parallelism</subject><subject>spike neural network</subject><subject>spike-based sparsity</subject><subject>System-on-chip</subject><subject>Training</subject><issn>1932-4545</issn><issn>1940-9990</issn><issn>1940-9990</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNpNkEtPAjEUhRujEUT_gDGmSzcDfc5M3SFB0RAhAdeTTnuRmnlgy0j894KAcXVucr9zFh9C15R0KSWqN38Y9GddRpjocpEQycgJalMlSKSUIqe7m7NISCFb6CKED0JkzBQ7Ry2uuIylEm1UD6eTwT3uY5ZGVYlll-PVS282meLhF1TryHq3TTzVXhcFFPgVGl-XtV8tncEj7e1Ge8Abt14eX7Yp9NrVVZTrABZPqsJVgMegfeWq90t0ttBFgKtDdtDb43A-GEXjydPzoD-ODGVERBY48JgnqZU2ToRkqRFK5TLPF9yYVCrIJV9oRkXMrIYYpJE6FTkx2uaQAu-gu_3uytefDYR1VrpgoCh0BXUTMk4pk0wmSmxRtkeNr0PwsMhW3pXaf2eUZDvR2a_obCc6O4jelm4P-01egv2rHM1ugZs94ADg32JCREwE_wHzvIF3</recordid><startdate>20241002</startdate><enddate>20241002</enddate><creator>Chen, Faquan</creator><creator>Tian, Qingyang</creator><creator>Xie, Lisheng</creator><creator>Zhou, Yifan</creator><creator>Wu, Ziren</creator><creator>Wu, Liangshun</creator><creator>Ying, Rendong</creator><creator>Wen, Fei</creator><creator>Liu, Peilin</creator><general>IEEE</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7X8</scope></search><sort><creationdate>20241002</creationdate><title>EPOC: A 28-nm 5.3 pJ/SOP Event-driven Parallel Neuromorphic Hardware with Neuromodulation-based Online Learning</title><author>Chen, Faquan ; Tian, Qingyang ; Xie, Lisheng ; Zhou, Yifan ; Wu, Ziren ; Wu, Liangshun ; Ying, Rendong ; Wen, Fei ; Liu, Peilin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c1204-de3e36378d5d674528c499b5bbf3cc859eb53fa21462dae6e5c5a84b0cadbe8e3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>artificial neural network</topic><topic>Biological neural networks</topic><topic>Complexity theory</topic><topic>Computational efficiency</topic><topic>Hardware</topic><topic>Memory management</topic><topic>neuromodulation</topic><topic>Neuromorphic hardware</topic><topic>Neuromorphics</topic><topic>Neurons</topic><topic>on-chip learning</topic><topic>Parallel processing</topic><topic>parallelism</topic><topic>spike neural network</topic><topic>spike-based sparsity</topic><topic>System-on-chip</topic><topic>Training</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Chen, Faquan</creatorcontrib><creatorcontrib>Tian, Qingyang</creatorcontrib><creatorcontrib>Xie, Lisheng</creatorcontrib><creatorcontrib>Zhou, Yifan</creatorcontrib><creatorcontrib>Wu, Ziren</creatorcontrib><creatorcontrib>Wu, Liangshun</creatorcontrib><creatorcontrib>Ying, Rendong</creatorcontrib><creatorcontrib>Wen, Fei</creatorcontrib><creatorcontrib>Liu, Peilin</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998-Present</collection><collection>IEEE/IET Electronic Library</collection><collection>PubMed</collection><collection>CrossRef</collection><collection>MEDLINE - Academic</collection><jtitle>IEEE transactions on biomedical circuits and systems</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Chen, Faquan</au><au>Tian, Qingyang</au><au>Xie, Lisheng</au><au>Zhou, Yifan</au><au>Wu, Ziren</au><au>Wu, Liangshun</au><au>Ying, Rendong</au><au>Wen, Fei</au><au>Liu, Peilin</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>EPOC: A 28-nm 5.3 pJ/SOP Event-driven Parallel Neuromorphic Hardware with Neuromodulation-based Online Learning</atitle><jtitle>IEEE transactions on biomedical circuits and systems</jtitle><stitle>TBCAS</stitle><addtitle>IEEE Trans Biomed Circuits Syst</addtitle><date>2024-10-02</date><risdate>2024</risdate><volume>PP</volume><spage>1</spage><epage>16</epage><pages>1-16</pages><issn>1932-4545</issn><issn>1940-9990</issn><eissn>1940-9990</eissn><coden>ITBCCW</coden><abstract>Bio-inspired neuromorphic hardware with learning ability is highly promising to achieve human-like intelligence, particularly in terms of high energy efficiency and strong environmental adaptability. Though many customized prototypes have demonstrated learning ability, learning on neuromorphic hardware still lacks a bio-plausible and unified learning framework, and inherent spike-based sparsity and parallelism have not been fully exploited, which fundamentally limits their computational efficiency and scale. Therefore, we develop a unified, event-driven, and massively parallel multi-core neuromorphic online learning processor, namely EPOC. We present a neuromodulation-based neuromorphic online learning framework to unify various learning algorithms, and EPOC supports high-accuracy local/global supervised Spike Neural Network (SNN) learning with a low-memory-demand streaming single-sample learning strategy through different neuromodulator formulations. EPOC leverages a novel event-driven computation method that fully exploits spike-based sparsity throughout the forward-backward learning phases, and parallel multi-channel and multi-core computing architecture, bringing 9.9× time efficiency improvement compared with the baseline architecture. We synthesize EPOC in a 28-nm CMOS process and perform extensive benchmarking. EPOC achieves state-of-the-art learning accuracy of 99.2%, 98.2%, and 94.3% on the MNIST, NMNIST, and DVS-Gesture benchmarks, respectively. Local-learning EPOC achieves 2.9× time efficiency improvement compared with the global learning counterpart. EPOC operates at a typical clock frequency of 100 MHz, providing a peak 328 GOPS/51 GSOPS throughput and a 5.3 pJ/SOP energy efficiency.</abstract><cop>United States</cop><pub>IEEE</pub><pmid>39356594</pmid><doi>10.1109/TBCAS.2024.3470520</doi><tpages>16</tpages></addata></record>
fulltext fulltext_linktorsrc
identifier ISSN: 1932-4545
ispartof IEEE transactions on biomedical circuits and systems, 2024-10, Vol.PP, p.1-16
issn 1932-4545
1940-9990
1940-9990
language eng
recordid cdi_ieee_primary_10704604
source IEEE/IET Electronic Library
subjects artificial neural network
Biological neural networks
Complexity theory
Computational efficiency
Hardware
Memory management
neuromodulation
Neuromorphic hardware
Neuromorphics
Neurons
on-chip learning
Parallel processing
parallelism
spike neural network
spike-based sparsity
System-on-chip
Training
title EPOC: A 28-nm 5.3 pJ/SOP Event-driven Parallel Neuromorphic Hardware with Neuromodulation-based Online Learning
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-07T14%3A46%3A33IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=EPOC:%20A%2028-nm%205.3%20pJ/SOP%20Event-driven%20Parallel%20Neuromorphic%20Hardware%20with%20Neuromodulation-based%20Online%20Learning&rft.jtitle=IEEE%20transactions%20on%20biomedical%20circuits%20and%20systems&rft.au=Chen,%20Faquan&rft.date=2024-10-02&rft.volume=PP&rft.spage=1&rft.epage=16&rft.pages=1-16&rft.issn=1932-4545&rft.eissn=1940-9990&rft.coden=ITBCCW&rft_id=info:doi/10.1109/TBCAS.2024.3470520&rft_dat=%3Cproquest_RIE%3E3112525794%3C/proquest_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=3112525794&rft_id=info:pmid/39356594&rft_ieee_id=10704604&rfr_iscdi=true