2Memristor‐1Capacitor Integrated Temporal Kernel for High‐Dimensional Data Mapping

Compact but precise feature‐extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high‐dimensional space using the time dynamics of a material system, such as a volatile m...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Small (Weinheim an der Bergstrasse, Germany) Germany), 2024-06, Vol.20 (25), p.e2306585-n/a
Hauptverfasser: Shim, Sung Keun, Jang, Yoon Ho, Han, Janguk, Jeon, Jeong Woo, Shin, Dong Hoon, Kim, Yeong Rok, Han, Joon‐Kyu, Woo, Kyung Seok, Lee, Soo Hyung, Cheong, Sunwoo, Kim, Jaehyun, Seo, Haengha, Shin, Jonghoon, Hwang, Cheol Seong
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page n/a
container_issue 25
container_start_page e2306585
container_title Small (Weinheim an der Bergstrasse, Germany)
container_volume 20
creator Shim, Sung Keun
Jang, Yoon Ho
Han, Janguk
Jeon, Jeong Woo
Shin, Dong Hoon
Kim, Yeong Rok
Han, Joon‐Kyu
Woo, Kyung Seok
Lee, Soo Hyung
Cheong, Sunwoo
Kim, Jaehyun
Seo, Haengha
Shin, Jonghoon
Hwang, Cheol Seong
description Compact but precise feature‐extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high‐dimensional space using the time dynamics of a material system, such as a volatile memristor. However, conventional physical RC systems have limited dynamics for the given material properties, restricting the methods to increase their dimensionality. This study proposes an integrated temporal kernel composed of a 2‐memristor and 1‐capacitor (2M1C) using a W/HfO2/TiN memristor and TiN/ZrO2/Al2O3/ZrO2/TiN capacitor to achieve higher dimensionality and tunable dynamics. The kernel elements are carefully designed and fabricated into an integrated array, of which performances are evaluated under diverse conditions. By optimizing the time dynamics of the 2M1C kernel, each memristor simultaneously extracts complementary information from input signals. The MNIST benchmark digit classification task achieves a high accuracy of 94.3% with a (196×10) single‐layer network. Analog input mapping ability is tested with a Mackey‐Glass time series prediction, and the system records a normalized root mean square error of 0.04 with a 20×1 readout network, the smallest readout network ever used for Mackey‐Glass prediction in RC. These performances demonstrate its high potential for efficient temporal data analysis. An integrated temporal kernel using two memristors and one capacitor is fabricated. This kernel extracts complementary features from the input, ultimately processing MNIST images at 8‐bit to achieve an accuracy of 94.3%. Excellent prediction performance for the Mackey‐Glass time series is verified with NRMSE of 0.04 in minimal network size (20 × 1).
doi_str_mv 10.1002/smll.202306585
format Article
fullrecord <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_miscellaneous_2929026830</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2929026830</sourcerecordid><originalsourceid>FETCH-LOGICAL-c3285-c98c74cd4d17a7d167045be90246205b283322b99b8cdfdb4c2c0b3145d384543</originalsourceid><addsrcrecordid>eNqFkLtOwzAUQC0EglJYGVEkFpYW-9pJ7BG1PCpSMfBYLcdxi6u8sBMhNj6Bb-RLMCoUiYXJtu65R9ZB6IjgMcEYznxVlmPAQHES83gLDUhC6CjhILY3d4L30L73K4wpAZbuoj3KgQBwMkCPMDeVs75r3MfbO5moVmkbHtGs7szSqc4U0b2p2sapMroxrjZltAjja7t8CgtTW5na26YO06nqVDRXbWvr5QHaWajSm8Pvc4geLi_uJ9ej7PZqNjnPRpoCj0dacJ0yXbCCpCotSJJiFudGYGAJ4DgHTilALkTOdbEocqZB45wSFheUs5jRITpde1vXPPfGd7KyXpuyVLVpei9BQJAlnOKAnvxBV03vwse9DPFETEXwBWq8prRrvHdmIVtnK-VeJcHyq7j8Ki43xcPC8be2zytTbPCfxAEQa-DFlub1H528m2fZr_wTTzeNqA</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3069539543</pqid></control><display><type>article</type><title>2Memristor‐1Capacitor Integrated Temporal Kernel for High‐Dimensional Data Mapping</title><source>Wiley-Blackwell Journals</source><creator>Shim, Sung Keun ; Jang, Yoon Ho ; Han, Janguk ; Jeon, Jeong Woo ; Shin, Dong Hoon ; Kim, Yeong Rok ; Han, Joon‐Kyu ; Woo, Kyung Seok ; Lee, Soo Hyung ; Cheong, Sunwoo ; Kim, Jaehyun ; Seo, Haengha ; Shin, Jonghoon ; Hwang, Cheol Seong</creator><creatorcontrib>Shim, Sung Keun ; Jang, Yoon Ho ; Han, Janguk ; Jeon, Jeong Woo ; Shin, Dong Hoon ; Kim, Yeong Rok ; Han, Joon‐Kyu ; Woo, Kyung Seok ; Lee, Soo Hyung ; Cheong, Sunwoo ; Kim, Jaehyun ; Seo, Haengha ; Shin, Jonghoon ; Hwang, Cheol Seong</creatorcontrib><description>Compact but precise feature‐extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high‐dimensional space using the time dynamics of a material system, such as a volatile memristor. However, conventional physical RC systems have limited dynamics for the given material properties, restricting the methods to increase their dimensionality. This study proposes an integrated temporal kernel composed of a 2‐memristor and 1‐capacitor (2M1C) using a W/HfO2/TiN memristor and TiN/ZrO2/Al2O3/ZrO2/TiN capacitor to achieve higher dimensionality and tunable dynamics. The kernel elements are carefully designed and fabricated into an integrated array, of which performances are evaluated under diverse conditions. By optimizing the time dynamics of the 2M1C kernel, each memristor simultaneously extracts complementary information from input signals. The MNIST benchmark digit classification task achieves a high accuracy of 94.3% with a (196×10) single‐layer network. Analog input mapping ability is tested with a Mackey‐Glass time series prediction, and the system records a normalized root mean square error of 0.04 with a 20×1 readout network, the smallest readout network ever used for Mackey‐Glass prediction in RC. These performances demonstrate its high potential for efficient temporal data analysis. An integrated temporal kernel using two memristors and one capacitor is fabricated. This kernel extracts complementary features from the input, ultimately processing MNIST images at 8‐bit to achieve an accuracy of 94.3%. Excellent prediction performance for the Mackey‐Glass time series is verified with NRMSE of 0.04 in minimal network size (20 × 1).</description><identifier>ISSN: 1613-6810</identifier><identifier>ISSN: 1613-6829</identifier><identifier>EISSN: 1613-6829</identifier><identifier>DOI: 10.1002/smll.202306585</identifier><identifier>PMID: 38212281</identifier><language>eng</language><publisher>Germany: Wiley Subscription Services, Inc</publisher><subject>Aluminum oxide ; analog memristor ; Capacitors ; Data analysis ; dual feature mapping ; Dynamics ; Mapping ; Material properties ; Memristors ; neuromorphic hardware ; Task complexity ; temporal data processing ; time series prediction ; Titanium nitride ; Zirconium dioxide</subject><ispartof>Small (Weinheim an der Bergstrasse, Germany), 2024-06, Vol.20 (25), p.e2306585-n/a</ispartof><rights>2024 Wiley‐VCH GmbH</rights><rights>2024 Wiley-VCH GmbH.</rights><rights>2024 Wiley‐VCH GmbH.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c3285-c98c74cd4d17a7d167045be90246205b283322b99b8cdfdb4c2c0b3145d384543</cites><orcidid>0009-0004-2071-2731 ; 0000-0002-6254-9758 ; 0009-0008-3718-1833</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://onlinelibrary.wiley.com/doi/pdf/10.1002%2Fsmll.202306585$$EPDF$$P50$$Gwiley$$H</linktopdf><linktohtml>$$Uhttps://onlinelibrary.wiley.com/doi/full/10.1002%2Fsmll.202306585$$EHTML$$P50$$Gwiley$$H</linktohtml><link.rule.ids>315,782,786,1419,27931,27932,45581,45582</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/38212281$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Shim, Sung Keun</creatorcontrib><creatorcontrib>Jang, Yoon Ho</creatorcontrib><creatorcontrib>Han, Janguk</creatorcontrib><creatorcontrib>Jeon, Jeong Woo</creatorcontrib><creatorcontrib>Shin, Dong Hoon</creatorcontrib><creatorcontrib>Kim, Yeong Rok</creatorcontrib><creatorcontrib>Han, Joon‐Kyu</creatorcontrib><creatorcontrib>Woo, Kyung Seok</creatorcontrib><creatorcontrib>Lee, Soo Hyung</creatorcontrib><creatorcontrib>Cheong, Sunwoo</creatorcontrib><creatorcontrib>Kim, Jaehyun</creatorcontrib><creatorcontrib>Seo, Haengha</creatorcontrib><creatorcontrib>Shin, Jonghoon</creatorcontrib><creatorcontrib>Hwang, Cheol Seong</creatorcontrib><title>2Memristor‐1Capacitor Integrated Temporal Kernel for High‐Dimensional Data Mapping</title><title>Small (Weinheim an der Bergstrasse, Germany)</title><addtitle>Small</addtitle><description>Compact but precise feature‐extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high‐dimensional space using the time dynamics of a material system, such as a volatile memristor. However, conventional physical RC systems have limited dynamics for the given material properties, restricting the methods to increase their dimensionality. This study proposes an integrated temporal kernel composed of a 2‐memristor and 1‐capacitor (2M1C) using a W/HfO2/TiN memristor and TiN/ZrO2/Al2O3/ZrO2/TiN capacitor to achieve higher dimensionality and tunable dynamics. The kernel elements are carefully designed and fabricated into an integrated array, of which performances are evaluated under diverse conditions. By optimizing the time dynamics of the 2M1C kernel, each memristor simultaneously extracts complementary information from input signals. The MNIST benchmark digit classification task achieves a high accuracy of 94.3% with a (196×10) single‐layer network. Analog input mapping ability is tested with a Mackey‐Glass time series prediction, and the system records a normalized root mean square error of 0.04 with a 20×1 readout network, the smallest readout network ever used for Mackey‐Glass prediction in RC. These performances demonstrate its high potential for efficient temporal data analysis. An integrated temporal kernel using two memristors and one capacitor is fabricated. This kernel extracts complementary features from the input, ultimately processing MNIST images at 8‐bit to achieve an accuracy of 94.3%. Excellent prediction performance for the Mackey‐Glass time series is verified with NRMSE of 0.04 in minimal network size (20 × 1).</description><subject>Aluminum oxide</subject><subject>analog memristor</subject><subject>Capacitors</subject><subject>Data analysis</subject><subject>dual feature mapping</subject><subject>Dynamics</subject><subject>Mapping</subject><subject>Material properties</subject><subject>Memristors</subject><subject>neuromorphic hardware</subject><subject>Task complexity</subject><subject>temporal data processing</subject><subject>time series prediction</subject><subject>Titanium nitride</subject><subject>Zirconium dioxide</subject><issn>1613-6810</issn><issn>1613-6829</issn><issn>1613-6829</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><recordid>eNqFkLtOwzAUQC0EglJYGVEkFpYW-9pJ7BG1PCpSMfBYLcdxi6u8sBMhNj6Bb-RLMCoUiYXJtu65R9ZB6IjgMcEYznxVlmPAQHES83gLDUhC6CjhILY3d4L30L73K4wpAZbuoj3KgQBwMkCPMDeVs75r3MfbO5moVmkbHtGs7szSqc4U0b2p2sapMroxrjZltAjja7t8CgtTW5na26YO06nqVDRXbWvr5QHaWajSm8Pvc4geLi_uJ9ej7PZqNjnPRpoCj0dacJ0yXbCCpCotSJJiFudGYGAJ4DgHTilALkTOdbEocqZB45wSFheUs5jRITpde1vXPPfGd7KyXpuyVLVpei9BQJAlnOKAnvxBV03vwse9DPFETEXwBWq8prRrvHdmIVtnK-VeJcHyq7j8Ki43xcPC8be2zytTbPCfxAEQa-DFlub1H528m2fZr_wTTzeNqA</recordid><startdate>20240601</startdate><enddate>20240601</enddate><creator>Shim, Sung Keun</creator><creator>Jang, Yoon Ho</creator><creator>Han, Janguk</creator><creator>Jeon, Jeong Woo</creator><creator>Shin, Dong Hoon</creator><creator>Kim, Yeong Rok</creator><creator>Han, Joon‐Kyu</creator><creator>Woo, Kyung Seok</creator><creator>Lee, Soo Hyung</creator><creator>Cheong, Sunwoo</creator><creator>Kim, Jaehyun</creator><creator>Seo, Haengha</creator><creator>Shin, Jonghoon</creator><creator>Hwang, Cheol Seong</creator><general>Wiley Subscription Services, Inc</general><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SR</scope><scope>7U5</scope><scope>8BQ</scope><scope>8FD</scope><scope>JG9</scope><scope>L7M</scope><scope>7X8</scope><orcidid>https://orcid.org/0009-0004-2071-2731</orcidid><orcidid>https://orcid.org/0000-0002-6254-9758</orcidid><orcidid>https://orcid.org/0009-0008-3718-1833</orcidid></search><sort><creationdate>20240601</creationdate><title>2Memristor‐1Capacitor Integrated Temporal Kernel for High‐Dimensional Data Mapping</title><author>Shim, Sung Keun ; Jang, Yoon Ho ; Han, Janguk ; Jeon, Jeong Woo ; Shin, Dong Hoon ; Kim, Yeong Rok ; Han, Joon‐Kyu ; Woo, Kyung Seok ; Lee, Soo Hyung ; Cheong, Sunwoo ; Kim, Jaehyun ; Seo, Haengha ; Shin, Jonghoon ; Hwang, Cheol Seong</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c3285-c98c74cd4d17a7d167045be90246205b283322b99b8cdfdb4c2c0b3145d384543</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Aluminum oxide</topic><topic>analog memristor</topic><topic>Capacitors</topic><topic>Data analysis</topic><topic>dual feature mapping</topic><topic>Dynamics</topic><topic>Mapping</topic><topic>Material properties</topic><topic>Memristors</topic><topic>neuromorphic hardware</topic><topic>Task complexity</topic><topic>temporal data processing</topic><topic>time series prediction</topic><topic>Titanium nitride</topic><topic>Zirconium dioxide</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Shim, Sung Keun</creatorcontrib><creatorcontrib>Jang, Yoon Ho</creatorcontrib><creatorcontrib>Han, Janguk</creatorcontrib><creatorcontrib>Jeon, Jeong Woo</creatorcontrib><creatorcontrib>Shin, Dong Hoon</creatorcontrib><creatorcontrib>Kim, Yeong Rok</creatorcontrib><creatorcontrib>Han, Joon‐Kyu</creatorcontrib><creatorcontrib>Woo, Kyung Seok</creatorcontrib><creatorcontrib>Lee, Soo Hyung</creatorcontrib><creatorcontrib>Cheong, Sunwoo</creatorcontrib><creatorcontrib>Kim, Jaehyun</creatorcontrib><creatorcontrib>Seo, Haengha</creatorcontrib><creatorcontrib>Shin, Jonghoon</creatorcontrib><creatorcontrib>Hwang, Cheol Seong</creatorcontrib><collection>PubMed</collection><collection>CrossRef</collection><collection>Engineered Materials Abstracts</collection><collection>Solid State and Superconductivity Abstracts</collection><collection>METADEX</collection><collection>Technology Research Database</collection><collection>Materials Research Database</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>MEDLINE - Academic</collection><jtitle>Small (Weinheim an der Bergstrasse, Germany)</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Shim, Sung Keun</au><au>Jang, Yoon Ho</au><au>Han, Janguk</au><au>Jeon, Jeong Woo</au><au>Shin, Dong Hoon</au><au>Kim, Yeong Rok</au><au>Han, Joon‐Kyu</au><au>Woo, Kyung Seok</au><au>Lee, Soo Hyung</au><au>Cheong, Sunwoo</au><au>Kim, Jaehyun</au><au>Seo, Haengha</au><au>Shin, Jonghoon</au><au>Hwang, Cheol Seong</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>2Memristor‐1Capacitor Integrated Temporal Kernel for High‐Dimensional Data Mapping</atitle><jtitle>Small (Weinheim an der Bergstrasse, Germany)</jtitle><addtitle>Small</addtitle><date>2024-06-01</date><risdate>2024</risdate><volume>20</volume><issue>25</issue><spage>e2306585</spage><epage>n/a</epage><pages>e2306585-n/a</pages><issn>1613-6810</issn><issn>1613-6829</issn><eissn>1613-6829</eissn><abstract>Compact but precise feature‐extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high‐dimensional space using the time dynamics of a material system, such as a volatile memristor. However, conventional physical RC systems have limited dynamics for the given material properties, restricting the methods to increase their dimensionality. This study proposes an integrated temporal kernel composed of a 2‐memristor and 1‐capacitor (2M1C) using a W/HfO2/TiN memristor and TiN/ZrO2/Al2O3/ZrO2/TiN capacitor to achieve higher dimensionality and tunable dynamics. The kernel elements are carefully designed and fabricated into an integrated array, of which performances are evaluated under diverse conditions. By optimizing the time dynamics of the 2M1C kernel, each memristor simultaneously extracts complementary information from input signals. The MNIST benchmark digit classification task achieves a high accuracy of 94.3% with a (196×10) single‐layer network. Analog input mapping ability is tested with a Mackey‐Glass time series prediction, and the system records a normalized root mean square error of 0.04 with a 20×1 readout network, the smallest readout network ever used for Mackey‐Glass prediction in RC. These performances demonstrate its high potential for efficient temporal data analysis. An integrated temporal kernel using two memristors and one capacitor is fabricated. This kernel extracts complementary features from the input, ultimately processing MNIST images at 8‐bit to achieve an accuracy of 94.3%. Excellent prediction performance for the Mackey‐Glass time series is verified with NRMSE of 0.04 in minimal network size (20 × 1).</abstract><cop>Germany</cop><pub>Wiley Subscription Services, Inc</pub><pmid>38212281</pmid><doi>10.1002/smll.202306585</doi><tpages>13</tpages><orcidid>https://orcid.org/0009-0004-2071-2731</orcidid><orcidid>https://orcid.org/0000-0002-6254-9758</orcidid><orcidid>https://orcid.org/0009-0008-3718-1833</orcidid></addata></record>
fulltext fulltext
identifier ISSN: 1613-6810
ispartof Small (Weinheim an der Bergstrasse, Germany), 2024-06, Vol.20 (25), p.e2306585-n/a
issn 1613-6810
1613-6829
1613-6829
language eng
recordid cdi_proquest_miscellaneous_2929026830
source Wiley-Blackwell Journals
subjects Aluminum oxide
analog memristor
Capacitors
Data analysis
dual feature mapping
Dynamics
Mapping
Material properties
Memristors
neuromorphic hardware
Task complexity
temporal data processing
time series prediction
Titanium nitride
Zirconium dioxide
title 2Memristor‐1Capacitor Integrated Temporal Kernel for High‐Dimensional Data Mapping
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-06T15%3A08%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=2Memristor%E2%80%901Capacitor%20Integrated%20Temporal%20Kernel%20for%20High%E2%80%90Dimensional%20Data%20Mapping&rft.jtitle=Small%20(Weinheim%20an%20der%20Bergstrasse,%20Germany)&rft.au=Shim,%20Sung%20Keun&rft.date=2024-06-01&rft.volume=20&rft.issue=25&rft.spage=e2306585&rft.epage=n/a&rft.pages=e2306585-n/a&rft.issn=1613-6810&rft.eissn=1613-6829&rft_id=info:doi/10.1002/smll.202306585&rft_dat=%3Cproquest_cross%3E2929026830%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=3069539543&rft_id=info:pmid/38212281&rfr_iscdi=true