Neural Architecture Search with In‐Memory Multiply–Accumulate and In‐Memory Rank Based on Coating Layer Optimized C‐Doped Ge2Sb2Te5 Phase Change Memory

Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design. However, the energy and time consumptions of conventional software‐based NAS are huge, hindering its development and applications. Herein, 4 Mb...

Celý popis

Uloženo v:

Podrobná bibliografie
Vydáno v:	Advanced functional materials Ročník 34; číslo 15
Hlavní autoři:	Yan, Longhao, Wu, Qingyu, Li, Xi, Xie, Chenchen, Zhou, Xilin, Li, Yuqi, Shi, Daijing, Yu, Lianfeng, Zhang, Teng, Tao, Yaoyu, Yan, Bonan, Zhong, Min, Song, Zhitang, Yang, Yuchao, Huang, Ru
Médium:	Journal Article
Jazyk:	angličtina
Vydáno:	Hoboken Wiley Subscription Services, Inc 10.04.2024
Témata:	Chips (memory devices) Cognitive tasks Computation Computer architecture Computer memory Graphics processing units in‐memory computing in‐memory rank Machine learning neural architecture search Neural networks operators phase change memory
ISSN:	1616-301X, 1616-3028
On-line přístup:	Získat plný text
Tagy:	Přidat tag Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!

Abstract	Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design. However, the energy and time consumptions of conventional software‐based NAS are huge, hindering its development and applications. Herein, 4 Mb phase change memory (PCM) chips are first fabricated that enable two key in‐memory computing operations—in‐memory multiply‐accumulate (MAC) and in‐memory rank for efficient NAS. The impacts of the coating layer material are systematically analyzed for the blade‐type heating electrode on the device uniformity and in turn NAS performance. The random weights in the searched network architecture can be fine‐tuned in the last stage. With 512 × 512 arrays based on 40 nm CMOS process, the PCM‐based NAS has achieved 25–53× smaller model size and better performance than manually designed networks and improved the energy and time efficiency by 4779× and 123×, respectively, compared with NAS running on graphic processing unit (GPU). This work can expand the hardware accelerated in‐memory operators, and significantly extend the applications of in‐memory computing enabled by nonvolatile memory in advanced machine learning tasks. Herein, 4 Mb phase change memory chips are first fabricated that enable two key in‐memory computing operations—in‐memory multiply accumulate (MAC) and in‐memory rank for efficient NAS. With 512 × 512 arrays the PCM‐based NAS has achieved 25–53× smaller model size and better performance than manually designed networks, and improved the energy and time efficiency by 4779× and 123×, respectively, compared with GPU.
AbstractList	Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design. However, the energy and time consumptions of conventional software‐based NAS are huge, hindering its development and applications. Herein, 4 Mb phase change memory (PCM) chips are first fabricated that enable two key in‐memory computing operations—in‐memory multiply‐accumulate (MAC) and in‐memory rank for efficient NAS. The impacts of the coating layer material are systematically analyzed for the blade‐type heating electrode on the device uniformity and in turn NAS performance. The random weights in the searched network architecture can be fine‐tuned in the last stage. With 512 × 512 arrays based on 40 nm CMOS process, the PCM‐based NAS has achieved 25–53× smaller model size and better performance than manually designed networks and improved the energy and time efficiency by 4779× and 123×, respectively, compared with NAS running on graphic processing unit (GPU). This work can expand the hardware accelerated in‐memory operators, and significantly extend the applications of in‐memory computing enabled by nonvolatile memory in advanced machine learning tasks. Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design. However, the energy and time consumptions of conventional software‐based NAS are huge, hindering its development and applications. Herein, 4 Mb phase change memory (PCM) chips are first fabricated that enable two key in‐memory computing operations—in‐memory multiply‐accumulate (MAC) and in‐memory rank for efficient NAS. The impacts of the coating layer material are systematically analyzed for the blade‐type heating electrode on the device uniformity and in turn NAS performance. The random weights in the searched network architecture can be fine‐tuned in the last stage. With 512 × 512 arrays based on 40 nm CMOS process, the PCM‐based NAS has achieved 25–53× smaller model size and better performance than manually designed networks and improved the energy and time efficiency by 4779× and 123×, respectively, compared with NAS running on graphic processing unit (GPU). This work can expand the hardware accelerated in‐memory operators, and significantly extend the applications of in‐memory computing enabled by nonvolatile memory in advanced machine learning tasks. Herein, 4 Mb phase change memory chips are first fabricated that enable two key in‐memory computing operations—in‐memory multiply accumulate (MAC) and in‐memory rank for efficient NAS. With 512 × 512 arrays the PCM‐based NAS has achieved 25–53× smaller model size and better performance than manually designed networks, and improved the energy and time efficiency by 4779× and 123×, respectively, compared with GPU.
Author	Yan, Longhao Zhang, Teng Tao, Yaoyu Zhong, Min Shi, Daijing Li, Xi Li, Yuqi Yan, Bonan Zhou, Xilin Yu, Lianfeng Wu, Qingyu Song, Zhitang Huang, Ru Xie, Chenchen Yang, Yuchao
Author_xml	– sequence: 1 givenname: Longhao surname: Yan fullname: Yan, Longhao organization: Peking University – sequence: 2 givenname: Qingyu surname: Wu fullname: Wu, Qingyu organization: Chinese Academy of Sciences – sequence: 3 givenname: Xi surname: Li fullname: Li, Xi organization: Chinese Academy of Sciences – sequence: 4 givenname: Chenchen surname: Xie fullname: Xie, Chenchen organization: Chinese Academy of Sciences – sequence: 5 givenname: Xilin surname: Zhou fullname: Zhou, Xilin organization: Chinese Academy of Sciences – sequence: 6 givenname: Yuqi surname: Li fullname: Li, Yuqi organization: Peking University – sequence: 7 givenname: Daijing surname: Shi fullname: Shi, Daijing organization: Peking University – sequence: 8 givenname: Lianfeng surname: Yu fullname: Yu, Lianfeng organization: Peking University – sequence: 9 givenname: Teng surname: Zhang fullname: Zhang, Teng organization: Peking University – sequence: 10 givenname: Yaoyu surname: Tao fullname: Tao, Yaoyu organization: Peking University – sequence: 11 givenname: Bonan surname: Yan fullname: Yan, Bonan organization: Peking University – sequence: 12 givenname: Min surname: Zhong fullname: Zhong, Min organization: Shanghai Integrated Circuit R&D Center – sequence: 13 givenname: Zhitang surname: Song fullname: Song, Zhitang email: ztsong@mail.sim.ac.cn organization: Chinese Academy of Sciences – sequence: 14 givenname: Yuchao orcidid: 0000-0003-4674-4059 surname: Yang fullname: Yang, Yuchao email: yuchaoyang@pku.edu.cn organization: Chinese Institute for Brain Research (CIBR) – sequence: 15 givenname: Ru surname: Huang fullname: Huang, Ru email: ruhuang@pku.edu.cn organization: Peking University
BookMark	eNpVUdtKw0AQXUTB66vPCz5H99a0eYypN2it2Aq-henutI0mm7hJKPGpnyD4Af5bv8SUSsF5mTnMOWdgzjHZt7lFQs45u-SMiSsws-xSMCEZU53eHjniPvc9yURvfzfz10NyXJZvjPFuV6oj8vOItYOUhk4vkgp1VTukY4QW0mVSLeiDXa--hpjlrqHDOq2SIm3Wq-9Q6zqrU6iQgjX_WM9g3-k1lGhobmmUQ5XYOR1Ag46OiirJks92FbWCfl600x2K8VRMsEOfFq2KRguwc6Rbt1NyMIO0xLO_fkJebm8m0b03GN09ROHAK4SUPU8oQMN7DNpCDVoa7uvNLwQL_A5KyRXowJipmnWNEgAm4GrKpK-7Wkw1yBNysfUtXP5RY1nFb3ntbHsylkyqQCjBZcsKtqxlkmITFy7JwDUxZ_EmgXiTQLxLIA77t8Mdkr-DLIOv
ContentType	Journal Article
Copyright	2023 Wiley‐VCH GmbH 2024 Wiley‐VCH GmbH
Copyright_xml	– notice: 2023 Wiley‐VCH GmbH – notice: 2024 Wiley‐VCH GmbH
DBID	7SP 7SR 7U5 8BQ 8FD JG9 L7M
DOI	10.1002/adfm.202300458
DatabaseName	Electronics & Communications Abstracts Engineered Materials Abstracts Solid State and Superconductivity Abstracts METADEX Technology Research Database Materials Research Database Advanced Technologies Database with Aerospace
DatabaseTitle	Materials Research Database Engineered Materials Abstracts Technology Research Database Electronics & Communications Abstracts Solid State and Superconductivity Abstracts Advanced Technologies Database with Aerospace METADEX
DatabaseTitleList	Materials Research Database
DeliveryMethod	fulltext_linktorsrc
Discipline	Engineering
EISSN	1616-3028
EndPage	n/a
ExternalDocumentID	ADFM202300458
Genre	article
GrantInformation_xml	– fundername: Youth Innovation Promotion Association, CAS funderid: 2018268 – fundername: National Natural Science Foundation of China funderid: 61925401; 92064004; 61927901; 92164302; 91964204 – fundername: Strategic Priority Research Program of the Chinese Academy of Sciences funderid: XDB44010200 – fundername: Science and Technology Council of Shanghai funderid: 19JC1416801 – fundername: 111 Project funderid: B18001
GroupedDBID	-~X .3N .GA 05W 0R~ 10A 1L6 1OC 23M 33P 3SF 3WU 4.4 4ZD 50Y 50Z 51W 51X 52M 52N 52O 52P 52S 52T 52U 52W 52X 53G 5GY 5VS 66C 6P2 702 7PT 8-0 8-1 8-3 8-4 8-5 8UM 930 A03 AAESR AAEVG AAHHS AAHQN AAMNL AANLZ AAONW AAXRX AAYCA AAZKR ABCQN ABCUV ABEML ABIJN ABJNI ABPVW ACAHQ ACCFJ ACCZN ACGFS ACIWK ACPOU ACSCC ACXBN ACXQS ADBBV ADEOM ADIZJ ADKYN ADMGS ADOZA ADXAS ADZMN ADZOD AEEZP AEIGN AEIMD AENEX AEQDE AEUQT AEUYR AFBPY AFFPM AFGKR AFPWT AFWVQ AFZJQ AHBTC AITYG AIURR AIWBW AJBDE AJXKR ALAGY ALMA_UNASSIGNED_HOLDINGS ALUQN ALVPJ AMBMR AMYDB ATUGU AUFTA AZBYB AZVAB BAFTC BDRZF BFHJK BHBCM BMNLL BMXJE BNHUX BROTX BRXPI BY8 CS3 D-E D-F DCZOG DPXWK DR2 DRFUL DRSTM EBS F00 F01 F04 F5P G-S G.N GNP GODZA H.T H.X HBH HGLYW HHY HHZ HZ~ IX1 J0M JPC KQQ LATKE LAW LC2 LC3 LEEKS LH4 LITHE LOXES LP6 LP7 LUTES LYRES MEWTI MK4 MRFUL MRSTM MSFUL MSSTM MXFUL MXSTM N04 N05 N9A NF~ NNB O66 O9- OIG P2P P2W P2X P4D Q.N Q11 QB0 QRW R.K RNS ROL RWI RX1 RYL SUPJJ UB1 V2E W8V W99 WBKPD WFSAM WIH WIK WJL WOHZO WQJ WRC WXSBR WYISQ XG1 XPP XV2 ~IA ~WT 7SP 7SR 7U5 8BQ 8FD AAMMB ADMLS AEFGJ AEYWJ AGHNM AGXDD AGYGG AIDQK AIDYY JG9 L7M
ID	FETCH-LOGICAL-p2338-24aed180aaaaecac3d16c202320965e3314ac9ddb4f7d42aad914b036c7c2bca3
IEDL.DBID	DRFUL
ISICitedReferencesCount	11
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001038798800001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN	1616-301X
IngestDate	Fri Jul 25 03:11:20 EDT 2025 Wed Jan 22 17:19:48 EST 2025
IsPeerReviewed	true
IsScholarly	true
Issue	15
Language	English
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-p2338-24aed180aaaaecac3d16c202320965e3314ac9ddb4f7d42aad914b036c7c2bca3
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ORCID	0000-0003-4674-4059
PQID	3034924213
PQPubID	2045204
PageCount	10
ParticipantIDs	proquest_journals_3034924213 wiley_primary_10_1002_adfm_202300458_ADFM202300458
PublicationCentury	2000
PublicationDate	April 10, 2024
PublicationDateYYYYMMDD	2024-04-10
PublicationDate_xml	– month: 04 year: 2024 text: April 10, 2024 day: 10
PublicationDecade	2020
PublicationPlace	Hoboken
PublicationPlace_xml	– name: Hoboken
PublicationTitle	Advanced functional materials
PublicationYear	2024
Publisher	Wiley Subscription Services, Inc
Publisher_xml	– name: Wiley Subscription Services, Inc
References	2022; 376 2021; 4 2020; 63 2019; 2 2019; 10 2002; 10 2020; 15 2019; 19 2007 2020; 11 2008; 1 2020; 6 2019; 20 2022 2022; 5 2018; 1 2021 2020 2022; 34 2019 2022; 13 2020; 577 2022; 57 2018 2017 2016 2022; 608 2022; 601
References_xml	– volume: 608 start-page: 504 year: 2022 publication-title: Nature – volume: 63 year: 2020 publication-title: Sci. China Inf. Sci. – volume: 5 start-page: 761 year: 2022 publication-title: Nat. Electron. – volume: 13 start-page: 3973 year: 2022 publication-title: Nat. Commun. – volume: 57 start-page: 1027 year: 2022 publication-title: IEEE J. Solid‐State Circuits – volume: 2 start-page: 290 year: 2019 publication-title: Nat. Electron. – volume: 376 start-page: 1066 year: 2022 publication-title: Science – year: 2007 – volume: 20 start-page: 1 year: 2019 publication-title: J. Mach. Learn. Res. – year: 2021 – volume: 577 start-page: 641 year: 2020 publication-title: Nature – volume: 6 year: 2020 publication-title: Sci. Adv. – volume: 601 start-page: 211 year: 2022 publication-title: Nature – volume: 10 start-page: 461 year: 2019 publication-title: Micromachines – year: 2016 – year: 2018 – volume: 34 year: 2022 publication-title: Adv. Mater. – volume: 11 start-page: 2473 year: 2020 publication-title: Nat. Commun. – volume: 13 start-page: 3765 year: 2022 publication-title: Nat. Commun. – volume: 1 start-page: 1313 year: 2008 publication-title: Proc. VLDB Endow. – volume: 1 start-page: 333 year: 2018 publication-title: Nat. Electron. – volume: 4 start-page: 81 year: 2021 publication-title: Nat. Electron. – year: 2022 – year: 2020 – volume: 4 start-page: 921 year: 2021 publication-title: Nat. Electron. – volume: 19 start-page: 164 year: 2019 publication-title: IEEE Trans. Device Mater. Reliab. – volume: 10 start-page: 99 year: 2002 publication-title: Evol. Comput. – year: 2017 – year: 2019 – volume: 15 start-page: 812 year: 2020 publication-title: Nat. Nanotechnol. – volume: 13 start-page: 4698 year: 2022 publication-title: Nat. Commun.
SSID	ssj0017734
Score	2.5027385
Snippet	Neural architecture search (NAS), as a subfield of automated machine learning, can design neural network models with better performance than manual design....
SourceID	proquest wiley
SourceType	Aggregation Database Publisher
SubjectTerms	Chips (memory devices) Cognitive tasks Computation Computer architecture Computer memory Graphics processing units in‐memory computing in‐memory rank Machine learning neural architecture search Neural networks operators phase change memory
Title	Neural Architecture Search with In‐Memory Multiply–Accumulate and In‐Memory Rank Based on Coating Layer Optimized C‐Doped Ge2Sb2Te5 Phase Change Memory
URI	https://onlinelibrary.wiley.com/doi/abs/10.1002%2Fadfm.202300458 https://www.proquest.com/docview/3034924213
Volume	34
WOSCitedRecordID	wos001038798800001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText	1
inHoldings	1
isFullTextHit
isPrint
journalDatabaseRights	– providerCode: PRVWIB databaseName: Wiley Online Library Full Collection 2020 customDbUrl: eissn: 1616-3028 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0017734 issn: 1616-301X databaseCode: DRFUL dateStart: 20010101 isFulltext: true titleUrlDefault: https://onlinelibrary.wiley.com providerName: Wiley-Blackwell
link	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV3JTsMwELWgcIADO6Js8oFrRL2kjo-lpYDUAmKReou8RVSItKKAVE58AhIfwL_xJYzjUtor5BA5isdKNDP2jGfmGaGDaqyN4pmMmCQ64o7ISAomIxWDu5VUndZZwemWOD9POh15OVHFH_AhxhtuXjOK-doruNKDw1_QUGUzX0lOPWRUnMyiOQrCy0tornHVvG2NIwlChMhylfgcL9L5AW6s0MPpEaZMzElDtVhpmsv__8YVtDSyMnEtiMUqmnH5GlqcwB5cR58elqPo8xtJwCH5GPvNWXyWf729t30i7hC3Q9rh8Ovto2bM84M_88thldupXlcqv8dHsCxa3Mtxvad8SjVuKbDq8QVMTQ_dV3hVB4JGrw-tE0evNb1xMb68AyocKh1wGG0D3TaPb-qn0ei0hqhPwc-NKFfOkqSi4HJGGWZJ1fh_px5gxjFGuDLSWs0zYTlVykrCNSygRhgKAsM2USnv5W4LYcEUyRzcpYi50VaDGyqyOGE8o0Cty2j3h1XpSOUGKfNIOz7AzcqIFkxJ-wGwIw3QzDT17EjH7EhrjWZ7_LT9F6IdtADtIpOHVHZR6enx2e2hefPy1B087o9E8Ru55OVI
linkProvider	Wiley-Blackwell
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV1LTxsxEB61oVLbA_RBVQq0PnBdET82Xh_ThADqJiAapNxWfq1AiE3EoxI98RMq9Qfw3_gleNZLINeqe1jtw2PtasaesefzZ4CtTmqsFqVKuKImEZ6qREmuEp2G4VbW8caUtaZzORplk4k6bNCEuBYm8kPMJ9ywZdT9NTZwnJDefmIN1a7EpeQMOaPS7CUsiWBLaQuW-keD43yeSpAyppY7FEFedPLI3Nhm24s1LMSYzyPV2tUMVv7DR76D5SbOJN1oGO_hha8-wNtn7IMf4Q6JOeoyT7kEEuHHBKdnyX51f_tniFDcGzKMwMOb-9u_XWuvz3HXL0905RZKHenqjHwPjtGRaUV6U42gapLrENeTg9A5nZ_-Dq96QaA_nYWrXc9-Gjb2KTk8CVIkrnUgsbZVOB7sjHt7SbNfQzJjYaSbMKG9o1lbh8NbbbmjHYv_zpBixnNOhbbKOSNK6QTT2ikqTHChVloWTIZ_glY1rfxnIJJrWvpwVjIV1jgTBqKyTDMuShakzRpsPOqqaBrdZcGRawdT3HwNWK2VYhYpO4pIzswKVEcxV0fR7Q-G87sv_yL0DV7vjYd5ke-PfqzDm_C8xvXQ9ga0ri6u_Sa8sr-uTi8vvjZ2-QBx_-k4
linkToPdf	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV1LTxsxELZaqBA9UFpa8Sr4wHVF_Nj1-hgSFhBJiChIua38WjWq2ESEVIITPwGJH8B_45cwXi-BXKvuYbUPj7Wr8XjGns-fEdpLYm0UL2TEJNERd0RGUjAZqRiGW2nitC4qTXdEr5cOBrJfown9WpjADzGbcPOWUfXX3sDd2Bb7b6yhyhZ-KTn1nFFx-hEt8lgmYJuL7fPssjNLJQgRUssJ8SAvMnhlbmzQ_fka5mLM95Fq5WqyL__hI1fRSh1n4mZoGF_RB1d-Q5_fsQ-uoSdPzFGVecsl4AA_xn56Fp-Uz_cPXQ_FvcXdADy8fb5_bBozvfK7fjmsSjtX6lyVf_ABOEaLRyVujZQHVeOOgrgen0HndDW8g1ctEGiPxnB15OgvTS9cjPu_QQqHtQ441PYdXWaHF63jqN6vIRpTGOlGlCtnSdpQcDijDLMkMf7fqaeYcYwRroy0VvNCWE6VspJwDS7UCEOhybAfaKEclW4dYcEUKRycpYi50VbDQFQUccp4QUFab6DtV13ltdFNcua5dnyKm20gWmklHwfKjjyQM9PcqyOfqSNvtrPu7G7zX4R20VK_neWdk97pFlqGxxWshzS20cLN9dT9RJ_M35vh5HqnbpYvH13osw
openUrl	ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Neural+Architecture+Search+with+In%E2%80%90Memory+Multiply%E2%80%93Accumulate+and+In%E2%80%90Memory+Rank+Based+on+Coating+Layer+Optimized+C%E2%80%90Doped+Ge2Sb2Te5+Phase+Change+Memory&rft.jtitle=Advanced+functional+materials&rft.au=Yan%2C+Longhao&rft.au=Wu%2C+Qingyu&rft.au=Li%2C+Xi&rft.au=Xie%2C+Chenchen&rft.date=2024-04-10&rft.pub=Wiley+Subscription+Services%2C+Inc&rft.issn=1616-301X&rft.eissn=1616-3028&rft.volume=34&rft.issue=15&rft_id=info:doi/10.1002%2Fadfm.202300458&rft.externalDBID=NO_FULL_TEXT
thumbnail_l	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1616-301X&client=summon
thumbnail_m	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1616-301X&client=summon
thumbnail_s	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1616-301X&client=summon