Excellent Pattern Recognition Accuracy of Neural Networks Using Hybrid Synapses and Complementary Training
To overcome the performance degradation in hardware neural networks (NNs) with non-ideal synapse devices, we proposed a novel neuromorphic architecture with both TiO x -based interfacial RRAM and CBRAM-based filamentary RRAM for highly accurate NN training and long-term inference reliability. We use...
Gespeichert in:
| Veröffentlicht in: | IEEE electron device letters 2021-04, Vol.42 (4), p.609-612 |
|---|---|
| Hauptverfasser: | , , , , , , |
| Format: | Artikel |
| Sprache: | eng |
| Schlagworte: | |
| Online-Zugang: | Volltext bestellen |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | 612 |
|---|---|
| container_issue | 4 |
| container_start_page | 609 |
| container_title | IEEE electron device letters |
| container_volume | 42 |
| creator | Kwak, Myonghoon Choi, Wooseok Heo, Seongjae Lee, Chuljun Nikam, Revannath Kim, Seyoung Hwang, Hyunsang |
| description | To overcome the performance degradation in hardware neural networks (NNs) with non-ideal synapse devices, we proposed a novel neuromorphic architecture with both TiO x -based interfacial RRAM and CBRAM-based filamentary RRAM for highly accurate NN training and long-term inference reliability. We used a threshold-triggered training scheme, in which interfacial and filamentary RRAMs were programmed in a complementary fashion. This took advantage of the long retention time of the filamentary RRAM and the high-resolution, symmetric weight update in the interfacial RRAM. Additional evaluation of device parameters, such as linearity, precision, variation, and retention time, was conducted. An excellent pattern recognition accuracy of ~97% was achieved during training with the MNIST dataset. Thus, reliable inference accuracy after training was maintained using the filamentary RRAM. |
| doi_str_mv | 10.1109/LED.2021.3058221 |
| format | Article |
| fullrecord | <record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_ieee_primary_9350640</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>9350640</ieee_id><sourcerecordid>2505613466</sourcerecordid><originalsourceid>FETCH-LOGICAL-c291t-c48472e1f9b94320e719999377164ff39880682b968cb0a1ae1bfb8023cb558f3</originalsourceid><addsrcrecordid>eNo9kM9PwjAUxxujiYjeTbw08Tx8r-227kgQxYSoUTgvXWnJcHTYjuj-e0sgvsv3Hb7f9-NDyC3CCBGKh_n0ccSA4YhDKhnDMzLANJUJpBk_JwPIBSYcIbskVyFsAFCIXAzIZvqrTdMY19F31XXGO_phdLt2dVe3jo613nule9pa-mpi20Tpflr_Fegy1G5NZ33l6xX97J3aBROocis6abe7xmzjUOV7uvCqdtF6TS6saoK5OemQLJ-mi8ksmb89v0zG80SzArtECylyZtAWVSE4A5NjEYvnOWbCWl5ICZlkVZFJXYFCZbCylQTGdRUftnxI7o9zd7793pvQlZt2711cWbI04kAusiy64OjSvg3BG1vufL2N95YI5YFoGYmWB6LliWiM3B0jtTHm317wFDIB_A8adnHK</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2505613466</pqid></control><display><type>article</type><title>Excellent Pattern Recognition Accuracy of Neural Networks Using Hybrid Synapses and Complementary Training</title><source>IEEE Electronic Library (IEL)</source><creator>Kwak, Myonghoon ; Choi, Wooseok ; Heo, Seongjae ; Lee, Chuljun ; Nikam, Revannath ; Kim, Seyoung ; Hwang, Hyunsang</creator><creatorcontrib>Kwak, Myonghoon ; Choi, Wooseok ; Heo, Seongjae ; Lee, Chuljun ; Nikam, Revannath ; Kim, Seyoung ; Hwang, Hyunsang</creatorcontrib><description>To overcome the performance degradation in hardware neural networks (NNs) with non-ideal synapse devices, we proposed a novel neuromorphic architecture with both TiO x -based interfacial RRAM and CBRAM-based filamentary RRAM for highly accurate NN training and long-term inference reliability. We used a threshold-triggered training scheme, in which interfacial and filamentary RRAMs were programmed in a complementary fashion. This took advantage of the long retention time of the filamentary RRAM and the high-resolution, symmetric weight update in the interfacial RRAM. Additional evaluation of device parameters, such as linearity, precision, variation, and retention time, was conducted. An excellent pattern recognition accuracy of ~97% was achieved during training with the MNIST dataset. Thus, reliable inference accuracy after training was maintained using the filamentary RRAM.</description><identifier>ISSN: 0741-3106</identifier><identifier>EISSN: 1558-0563</identifier><identifier>DOI: 10.1109/LED.2021.3058221</identifier><identifier>CODEN: EDLEDZ</identifier><language>eng</language><publisher>New York: IEEE</publisher><subject>Accuracy ; Artificial neural networks ; Crossbar array ; Hardware ; Inference ; Linearity ; Neural networks ; Object recognition ; Pattern recognition ; Performance degradation ; Performance evaluation ; synapse device ; Synapses ; Titanium oxides ; Training</subject><ispartof>IEEE electron device letters, 2021-04, Vol.42 (4), p.609-612</ispartof><rights>Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2021</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c291t-c48472e1f9b94320e719999377164ff39880682b968cb0a1ae1bfb8023cb558f3</citedby><cites>FETCH-LOGICAL-c291t-c48472e1f9b94320e719999377164ff39880682b968cb0a1ae1bfb8023cb558f3</cites><orcidid>0000-0003-1930-1914 ; 0000-0003-4064-4575 ; 0000-0002-7190-6980 ; 0000-0002-0786-0885 ; 0000-0002-0408-1165</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/9350640$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,777,781,793,27905,27906,54739</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/9350640$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Kwak, Myonghoon</creatorcontrib><creatorcontrib>Choi, Wooseok</creatorcontrib><creatorcontrib>Heo, Seongjae</creatorcontrib><creatorcontrib>Lee, Chuljun</creatorcontrib><creatorcontrib>Nikam, Revannath</creatorcontrib><creatorcontrib>Kim, Seyoung</creatorcontrib><creatorcontrib>Hwang, Hyunsang</creatorcontrib><title>Excellent Pattern Recognition Accuracy of Neural Networks Using Hybrid Synapses and Complementary Training</title><title>IEEE electron device letters</title><addtitle>LED</addtitle><description>To overcome the performance degradation in hardware neural networks (NNs) with non-ideal synapse devices, we proposed a novel neuromorphic architecture with both TiO x -based interfacial RRAM and CBRAM-based filamentary RRAM for highly accurate NN training and long-term inference reliability. We used a threshold-triggered training scheme, in which interfacial and filamentary RRAMs were programmed in a complementary fashion. This took advantage of the long retention time of the filamentary RRAM and the high-resolution, symmetric weight update in the interfacial RRAM. Additional evaluation of device parameters, such as linearity, precision, variation, and retention time, was conducted. An excellent pattern recognition accuracy of ~97% was achieved during training with the MNIST dataset. Thus, reliable inference accuracy after training was maintained using the filamentary RRAM.</description><subject>Accuracy</subject><subject>Artificial neural networks</subject><subject>Crossbar array</subject><subject>Hardware</subject><subject>Inference</subject><subject>Linearity</subject><subject>Neural networks</subject><subject>Object recognition</subject><subject>Pattern recognition</subject><subject>Performance degradation</subject><subject>Performance evaluation</subject><subject>synapse device</subject><subject>Synapses</subject><subject>Titanium oxides</subject><subject>Training</subject><issn>0741-3106</issn><issn>1558-0563</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNo9kM9PwjAUxxujiYjeTbw08Tx8r-227kgQxYSoUTgvXWnJcHTYjuj-e0sgvsv3Hb7f9-NDyC3CCBGKh_n0ccSA4YhDKhnDMzLANJUJpBk_JwPIBSYcIbskVyFsAFCIXAzIZvqrTdMY19F31XXGO_phdLt2dVe3jo613nule9pa-mpi20Tpflr_Fegy1G5NZ33l6xX97J3aBROocis6abe7xmzjUOV7uvCqdtF6TS6saoK5OemQLJ-mi8ksmb89v0zG80SzArtECylyZtAWVSE4A5NjEYvnOWbCWl5ICZlkVZFJXYFCZbCylQTGdRUftnxI7o9zd7793pvQlZt2711cWbI04kAusiy64OjSvg3BG1vufL2N95YI5YFoGYmWB6LliWiM3B0jtTHm317wFDIB_A8adnHK</recordid><startdate>20210401</startdate><enddate>20210401</enddate><creator>Kwak, Myonghoon</creator><creator>Choi, Wooseok</creator><creator>Heo, Seongjae</creator><creator>Lee, Chuljun</creator><creator>Nikam, Revannath</creator><creator>Kim, Seyoung</creator><creator>Hwang, Hyunsang</creator><general>IEEE</general><general>The Institute of Electrical and Electronics Engineers, Inc. (IEEE)</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7SP</scope><scope>8FD</scope><scope>L7M</scope><orcidid>https://orcid.org/0000-0003-1930-1914</orcidid><orcidid>https://orcid.org/0000-0003-4064-4575</orcidid><orcidid>https://orcid.org/0000-0002-7190-6980</orcidid><orcidid>https://orcid.org/0000-0002-0786-0885</orcidid><orcidid>https://orcid.org/0000-0002-0408-1165</orcidid></search><sort><creationdate>20210401</creationdate><title>Excellent Pattern Recognition Accuracy of Neural Networks Using Hybrid Synapses and Complementary Training</title><author>Kwak, Myonghoon ; Choi, Wooseok ; Heo, Seongjae ; Lee, Chuljun ; Nikam, Revannath ; Kim, Seyoung ; Hwang, Hyunsang</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c291t-c48472e1f9b94320e719999377164ff39880682b968cb0a1ae1bfb8023cb558f3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Accuracy</topic><topic>Artificial neural networks</topic><topic>Crossbar array</topic><topic>Hardware</topic><topic>Inference</topic><topic>Linearity</topic><topic>Neural networks</topic><topic>Object recognition</topic><topic>Pattern recognition</topic><topic>Performance degradation</topic><topic>Performance evaluation</topic><topic>synapse device</topic><topic>Synapses</topic><topic>Titanium oxides</topic><topic>Training</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Kwak, Myonghoon</creatorcontrib><creatorcontrib>Choi, Wooseok</creatorcontrib><creatorcontrib>Heo, Seongjae</creatorcontrib><creatorcontrib>Lee, Chuljun</creatorcontrib><creatorcontrib>Nikam, Revannath</creatorcontrib><creatorcontrib>Kim, Seyoung</creatorcontrib><creatorcontrib>Hwang, Hyunsang</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998-Present</collection><collection>IEEE Electronic Library (IEL)</collection><collection>CrossRef</collection><collection>Electronics & Communications Abstracts</collection><collection>Technology Research Database</collection><collection>Advanced Technologies Database with Aerospace</collection><jtitle>IEEE electron device letters</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kwak, Myonghoon</au><au>Choi, Wooseok</au><au>Heo, Seongjae</au><au>Lee, Chuljun</au><au>Nikam, Revannath</au><au>Kim, Seyoung</au><au>Hwang, Hyunsang</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Excellent Pattern Recognition Accuracy of Neural Networks Using Hybrid Synapses and Complementary Training</atitle><jtitle>IEEE electron device letters</jtitle><stitle>LED</stitle><date>2021-04-01</date><risdate>2021</risdate><volume>42</volume><issue>4</issue><spage>609</spage><epage>612</epage><pages>609-612</pages><issn>0741-3106</issn><eissn>1558-0563</eissn><coden>EDLEDZ</coden><abstract>To overcome the performance degradation in hardware neural networks (NNs) with non-ideal synapse devices, we proposed a novel neuromorphic architecture with both TiO x -based interfacial RRAM and CBRAM-based filamentary RRAM for highly accurate NN training and long-term inference reliability. We used a threshold-triggered training scheme, in which interfacial and filamentary RRAMs were programmed in a complementary fashion. This took advantage of the long retention time of the filamentary RRAM and the high-resolution, symmetric weight update in the interfacial RRAM. Additional evaluation of device parameters, such as linearity, precision, variation, and retention time, was conducted. An excellent pattern recognition accuracy of ~97% was achieved during training with the MNIST dataset. Thus, reliable inference accuracy after training was maintained using the filamentary RRAM.</abstract><cop>New York</cop><pub>IEEE</pub><doi>10.1109/LED.2021.3058221</doi><tpages>4</tpages><orcidid>https://orcid.org/0000-0003-1930-1914</orcidid><orcidid>https://orcid.org/0000-0003-4064-4575</orcidid><orcidid>https://orcid.org/0000-0002-7190-6980</orcidid><orcidid>https://orcid.org/0000-0002-0786-0885</orcidid><orcidid>https://orcid.org/0000-0002-0408-1165</orcidid></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | ISSN: 0741-3106 |
| ispartof | IEEE electron device letters, 2021-04, Vol.42 (4), p.609-612 |
| issn | 0741-3106 1558-0563 |
| language | eng |
| recordid | cdi_ieee_primary_9350640 |
| source | IEEE Electronic Library (IEL) |
| subjects | Accuracy Artificial neural networks Crossbar array Hardware Inference Linearity Neural networks Object recognition Pattern recognition Performance degradation Performance evaluation synapse device Synapses Titanium oxides Training |
| title | Excellent Pattern Recognition Accuracy of Neural Networks Using Hybrid Synapses and Complementary Training |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-19T17%3A15%3A52IST&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=Excellent%20Pattern%20Recognition%20Accuracy%20of%20Neural%20Networks%20Using%20Hybrid%20Synapses%20and%20Complementary%20Training&rft.jtitle=IEEE%20electron%20device%20letters&rft.au=Kwak,%20Myonghoon&rft.date=2021-04-01&rft.volume=42&rft.issue=4&rft.spage=609&rft.epage=612&rft.pages=609-612&rft.issn=0741-3106&rft.eissn=1558-0563&rft.coden=EDLEDZ&rft_id=info:doi/10.1109/LED.2021.3058221&rft_dat=%3Cproquest_RIE%3E2505613466%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=2505613466&rft_id=info:pmid/&rft_ieee_id=9350640&rfr_iscdi=true |