Calibration of Compact Resist Model Through CNN Training
We notice that the compact resist model can be mapped to a simple CNN (convolutional neural network): convolutional layer corresponds to convolutions between input images and resist kernels, and a fully connected layer can model the formation of weighted sum of convolutions followed by the compariso...
Gespeichert in:
Veröffentlicht in: | IEEE transactions on semiconductor manufacturing 2023-05, Vol.36 (2), p.1-1 |
---|---|
Hauptverfasser: | , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | 1 |
---|---|
container_issue | 2 |
container_start_page | 1 |
container_title | IEEE transactions on semiconductor manufacturing |
container_volume | 36 |
creator | Kwon, Yonghwi Shin, Youngsoo |
description | We notice that the compact resist model can be mapped to a simple CNN (convolutional neural network): convolutional layer corresponds to convolutions between input images and resist kernels, and a fully connected layer can model the formation of weighted sum of convolutions followed by the comparison to threshold to determine the development. Resist kernels correspond to convolution filters, so they can be obtained through CNN training, which is a key motivation. A number of challenges are identified and solutions are proposed: (1) We demonstrate a CNN structure that can be mapped to a resist model. (2) Convolution filters are large images and cannot be trained with standard methods. Adaptive learning rate and gradient clipping are applied. (3) Convolution filters may easily be overfitted if training data is not enough. We apply gradient descent for fast initialization of filters. (4) CNN is trained with printability of image pixels rather than CD values. Extraction of pixels and their sampling are addressed. The number of kernels and the number of convolutions can greatly be reduced through the proposed method: 22 standard kernel functions are reduced to only 4 optimized ones, which contributes to 44% faster lithography simulation yet accuracy is improved by 10%. |
doi_str_mv | 10.1109/TSM.2023.3267670 |
format | Article |
fullrecord | <record><control><sourceid>proquest_RIE</sourceid><recordid>TN_cdi_ieee_primary_10103175</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10103175</ieee_id><sourcerecordid>2809879708</sourcerecordid><originalsourceid>FETCH-LOGICAL-c245t-b4ae885974ead74b63e3b8324cfeae342a8fc9080270042e9ed27a2b82800f643</originalsourceid><addsrcrecordid>eNpNkD1PwzAURS0EEqWwMzBYYk55_ortEUVQkNoiQZgtJ31pU7VxsdOBf0-qdmC6y7n3SoeQewYTxsA-lV_zCQcuJoLnOtdwQUZMKZNxIdUlGYGxMssV6Gtyk9IGgElp9YiYwm_bKvq-DR0NDS3Cbu_rnn5ialNP52GJW1quYzis1rRYLGgZfdu13eqWXDV-m_DunGPy_fpSFm_Z7GP6XjzPsppL1WeV9GiMslqiX2pZ5QJFZQSXdYMeheTeNLUFA1wDSI4Wl1x7XhluAJpcijF5PO3uY_g5YOrdJhxiN1y6AbFGWw1moOBE1TGkFLFx-9jufPx1DNzRjxv8uKMfd_YzVB5OlRYR_-EMBNNK_AHItl9X</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2809879708</pqid></control><display><type>article</type><title>Calibration of Compact Resist Model Through CNN Training</title><source>IEEE Electronic Library (IEL)</source><creator>Kwon, Yonghwi ; Shin, Youngsoo</creator><creatorcontrib>Kwon, Yonghwi ; Shin, Youngsoo</creatorcontrib><description>We notice that the compact resist model can be mapped to a simple CNN (convolutional neural network): convolutional layer corresponds to convolutions between input images and resist kernels, and a fully connected layer can model the formation of weighted sum of convolutions followed by the comparison to threshold to determine the development. Resist kernels correspond to convolution filters, so they can be obtained through CNN training, which is a key motivation. A number of challenges are identified and solutions are proposed: (1) We demonstrate a CNN structure that can be mapped to a resist model. (2) Convolution filters are large images and cannot be trained with standard methods. Adaptive learning rate and gradient clipping are applied. (3) Convolution filters may easily be overfitted if training data is not enough. We apply gradient descent for fast initialization of filters. (4) CNN is trained with printability of image pixels rather than CD values. Extraction of pixels and their sampling are addressed. The number of kernels and the number of convolutions can greatly be reduced through the proposed method: 22 standard kernel functions are reduced to only 4 optimized ones, which contributes to 44% faster lithography simulation yet accuracy is improved by 10%.</description><identifier>ISSN: 0894-6507</identifier><identifier>EISSN: 1558-2345</identifier><identifier>DOI: 10.1109/TSM.2023.3267670</identifier><identifier>CODEN: ITSMED</identifier><language>eng</language><publisher>New York: IEEE</publisher><subject>Artificial neural networks ; Compact resist model ; Computational modeling ; Convolution ; convolutional neural network (CNN) ; Convolutional neural networks ; Image filters ; Kernel ; Kernel functions ; Pixels ; resist kernels ; Resists ; Semiconductor device modeling ; Training</subject><ispartof>IEEE transactions on semiconductor manufacturing, 2023-05, Vol.36 (2), p.1-1</ispartof><rights>Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c245t-b4ae885974ead74b63e3b8324cfeae342a8fc9080270042e9ed27a2b82800f643</cites><orcidid>0000-0001-5902-7900</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10103175$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,780,784,796,27924,27925,54758</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10103175$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Kwon, Yonghwi</creatorcontrib><creatorcontrib>Shin, Youngsoo</creatorcontrib><title>Calibration of Compact Resist Model Through CNN Training</title><title>IEEE transactions on semiconductor manufacturing</title><addtitle>TSM</addtitle><description>We notice that the compact resist model can be mapped to a simple CNN (convolutional neural network): convolutional layer corresponds to convolutions between input images and resist kernels, and a fully connected layer can model the formation of weighted sum of convolutions followed by the comparison to threshold to determine the development. Resist kernels correspond to convolution filters, so they can be obtained through CNN training, which is a key motivation. A number of challenges are identified and solutions are proposed: (1) We demonstrate a CNN structure that can be mapped to a resist model. (2) Convolution filters are large images and cannot be trained with standard methods. Adaptive learning rate and gradient clipping are applied. (3) Convolution filters may easily be overfitted if training data is not enough. We apply gradient descent for fast initialization of filters. (4) CNN is trained with printability of image pixels rather than CD values. Extraction of pixels and their sampling are addressed. The number of kernels and the number of convolutions can greatly be reduced through the proposed method: 22 standard kernel functions are reduced to only 4 optimized ones, which contributes to 44% faster lithography simulation yet accuracy is improved by 10%.</description><subject>Artificial neural networks</subject><subject>Compact resist model</subject><subject>Computational modeling</subject><subject>Convolution</subject><subject>convolutional neural network (CNN)</subject><subject>Convolutional neural networks</subject><subject>Image filters</subject><subject>Kernel</subject><subject>Kernel functions</subject><subject>Pixels</subject><subject>resist kernels</subject><subject>Resists</subject><subject>Semiconductor device modeling</subject><subject>Training</subject><issn>0894-6507</issn><issn>1558-2345</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNpNkD1PwzAURS0EEqWwMzBYYk55_ortEUVQkNoiQZgtJ31pU7VxsdOBf0-qdmC6y7n3SoeQewYTxsA-lV_zCQcuJoLnOtdwQUZMKZNxIdUlGYGxMssV6Gtyk9IGgElp9YiYwm_bKvq-DR0NDS3Cbu_rnn5ialNP52GJW1quYzis1rRYLGgZfdu13eqWXDV-m_DunGPy_fpSFm_Z7GP6XjzPsppL1WeV9GiMslqiX2pZ5QJFZQSXdYMeheTeNLUFA1wDSI4Wl1x7XhluAJpcijF5PO3uY_g5YOrdJhxiN1y6AbFGWw1moOBE1TGkFLFx-9jufPx1DNzRjxv8uKMfd_YzVB5OlRYR_-EMBNNK_AHItl9X</recordid><startdate>20230501</startdate><enddate>20230501</enddate><creator>Kwon, Yonghwi</creator><creator>Shin, Youngsoo</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>7U5</scope><scope>8FD</scope><scope>L7M</scope><orcidid>https://orcid.org/0000-0001-5902-7900</orcidid></search><sort><creationdate>20230501</creationdate><title>Calibration of Compact Resist Model Through CNN Training</title><author>Kwon, Yonghwi ; Shin, Youngsoo</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c245t-b4ae885974ead74b63e3b8324cfeae342a8fc9080270042e9ed27a2b82800f643</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Artificial neural networks</topic><topic>Compact resist model</topic><topic>Computational modeling</topic><topic>Convolution</topic><topic>convolutional neural network (CNN)</topic><topic>Convolutional neural networks</topic><topic>Image filters</topic><topic>Kernel</topic><topic>Kernel functions</topic><topic>Pixels</topic><topic>resist kernels</topic><topic>Resists</topic><topic>Semiconductor device modeling</topic><topic>Training</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Kwon, Yonghwi</creatorcontrib><creatorcontrib>Shin, Youngsoo</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>Solid State and Superconductivity Abstracts</collection><collection>Technology Research Database</collection><collection>Advanced Technologies Database with Aerospace</collection><jtitle>IEEE transactions on semiconductor manufacturing</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Kwon, Yonghwi</au><au>Shin, Youngsoo</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Calibration of Compact Resist Model Through CNN Training</atitle><jtitle>IEEE transactions on semiconductor manufacturing</jtitle><stitle>TSM</stitle><date>2023-05-01</date><risdate>2023</risdate><volume>36</volume><issue>2</issue><spage>1</spage><epage>1</epage><pages>1-1</pages><issn>0894-6507</issn><eissn>1558-2345</eissn><coden>ITSMED</coden><abstract>We notice that the compact resist model can be mapped to a simple CNN (convolutional neural network): convolutional layer corresponds to convolutions between input images and resist kernels, and a fully connected layer can model the formation of weighted sum of convolutions followed by the comparison to threshold to determine the development. Resist kernels correspond to convolution filters, so they can be obtained through CNN training, which is a key motivation. A number of challenges are identified and solutions are proposed: (1) We demonstrate a CNN structure that can be mapped to a resist model. (2) Convolution filters are large images and cannot be trained with standard methods. Adaptive learning rate and gradient clipping are applied. (3) Convolution filters may easily be overfitted if training data is not enough. We apply gradient descent for fast initialization of filters. (4) CNN is trained with printability of image pixels rather than CD values. Extraction of pixels and their sampling are addressed. The number of kernels and the number of convolutions can greatly be reduced through the proposed method: 22 standard kernel functions are reduced to only 4 optimized ones, which contributes to 44% faster lithography simulation yet accuracy is improved by 10%.</abstract><cop>New York</cop><pub>IEEE</pub><doi>10.1109/TSM.2023.3267670</doi><tpages>1</tpages><orcidid>https://orcid.org/0000-0001-5902-7900</orcidid></addata></record> |
fulltext | fulltext_linktorsrc |
identifier | ISSN: 0894-6507 |
ispartof | IEEE transactions on semiconductor manufacturing, 2023-05, Vol.36 (2), p.1-1 |
issn | 0894-6507 1558-2345 |
language | eng |
recordid | cdi_ieee_primary_10103175 |
source | IEEE Electronic Library (IEL) |
subjects | Artificial neural networks Compact resist model Computational modeling Convolution convolutional neural network (CNN) Convolutional neural networks Image filters Kernel Kernel functions Pixels resist kernels Resists Semiconductor device modeling Training |
title | Calibration of Compact Resist Model Through CNN Training |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-25T11%3A27%3A47IST&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=Calibration%20of%20Compact%20Resist%20Model%20Through%20CNN%20Training&rft.jtitle=IEEE%20transactions%20on%20semiconductor%20manufacturing&rft.au=Kwon,%20Yonghwi&rft.date=2023-05-01&rft.volume=36&rft.issue=2&rft.spage=1&rft.epage=1&rft.pages=1-1&rft.issn=0894-6507&rft.eissn=1558-2345&rft.coden=ITSMED&rft_id=info:doi/10.1109/TSM.2023.3267670&rft_dat=%3Cproquest_RIE%3E2809879708%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=2809879708&rft_id=info:pmid/&rft_ieee_id=10103175&rfr_iscdi=true |