Improving Deep Neural Networks' Training for Image Classification With Nonlinear Conjugate Gradient-Style Adaptive Momentum
Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning practice, the momentum is usually weighted by a well-calibrated constant. However, tuning the hyperparameter for momentum can be a significant...
Uloženo v:
| Vydáno v: | IEEE transaction on neural networks and learning systems Ročník 35; číslo 9; s. 12288 - 12300 |
|---|---|
| Hlavní autoři: | , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
United States
IEEE
01.09.2024
|
| Témata: | |
| ISSN: | 2162-237X, 2162-2388, 2162-2388 |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Abstract | Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning practice, the momentum is usually weighted by a well-calibrated constant. However, tuning the hyperparameter for momentum can be a significant computational burden. In this article, we propose a novel adaptive momentum for improving DNNs training; this adaptive momentum, with no momentum-related hyperparameter required, is motivated by the nonlinear conjugate gradient (NCG) method. Stochastic gradient descent (SGD) with this new adaptive momentum eliminates the need for the momentum hyperparameter calibration, allows using a significantly larger learning rate, accelerates DNN training, and improves the final accuracy and robustness of the trained DNNs. For instance, SGD with this adaptive momentum reduces classification errors for training ResNet110 for CIFAR10 and CIFAR100 from 5.25% to 4.64% and 23.75% to 20.03%, respectively. Furthermore, SGD, with the new adaptive momentum, also benefits adversarial training and, hence, improves the adversarial robustness of the trained DNNs. |
|---|---|
| AbstractList | Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning practice, the momentum is usually weighted by a well-calibrated constant. However, tuning the hyperparameter for momentum can be a significant computational burden. In this article, we propose a novel adaptive momentum for improving DNNs training; this adaptive momentum, with no momentum-related hyperparameter required, is motivated by the nonlinear conjugate gradient (NCG) method. Stochastic gradient descent (SGD) with this new adaptive momentum eliminates the need for the momentum hyperparameter calibration, allows using a significantly larger learning rate, accelerates DNN training, and improves the final accuracy and robustness of the trained DNNs. For instance, SGD with this adaptive momentum reduces classification errors for training ResNet110 for CIFAR10 and CIFAR100 from 5.25% to 4.64% and 23.75% to 20.03%, respectively. Furthermore, SGD, with the new adaptive momentum, also benefits adversarial training and, hence, improves the adversarial robustness of the trained DNNs.Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning practice, the momentum is usually weighted by a well-calibrated constant. However, tuning the hyperparameter for momentum can be a significant computational burden. In this article, we propose a novel adaptive momentum for improving DNNs training; this adaptive momentum, with no momentum-related hyperparameter required, is motivated by the nonlinear conjugate gradient (NCG) method. Stochastic gradient descent (SGD) with this new adaptive momentum eliminates the need for the momentum hyperparameter calibration, allows using a significantly larger learning rate, accelerates DNN training, and improves the final accuracy and robustness of the trained DNNs. For instance, SGD with this adaptive momentum reduces classification errors for training ResNet110 for CIFAR10 and CIFAR100 from 5.25% to 4.64% and 23.75% to 20.03%, respectively. Furthermore, SGD, with the new adaptive momentum, also benefits adversarial training and, hence, improves the adversarial robustness of the trained DNNs. Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning practice, the momentum is usually weighted by a well-calibrated constant. However, tuning the hyperparameter for momentum can be a significant computational burden. In this article, we propose a novel adaptive momentum for improving DNNs training; this adaptive momentum, with no momentum-related hyperparameter required, is motivated by the nonlinear conjugate gradient (NCG) method. Stochastic gradient descent (SGD) with this new adaptive momentum eliminates the need for the momentum hyperparameter calibration, allows using a significantly larger learning rate, accelerates DNN training, and improves the final accuracy and robustness of the trained DNNs. For instance, SGD with this adaptive momentum reduces classification errors for training ResNet110 for CIFAR10 and CIFAR100 from 5.25% to 4.64% and 23.75% to 20.03%, respectively. Furthermore, SGD, with the new adaptive momentum, also benefits adversarial training and, hence, improves the adversarial robustness of the trained DNNs. |
| Author | Wang, Bao Ye, Qiang |
| Author_xml | – sequence: 1 givenname: Bao orcidid: 0000-0002-4848-4791 surname: Wang fullname: Wang, Bao email: wangbaonj@gmail.com organization: Department of Mathematics and Scientific Computing and Imaging Institute, The University of Utah, Salt Lake City, UT, USA – sequence: 2 givenname: Qiang surname: Ye fullname: Ye, Qiang email: qye3@uky.edu organization: Department of Mathematics, University of Kentucky, Lexington, KY, USA |
| BackLink | https://www.ncbi.nlm.nih.gov/pubmed/37030680$$D View this record in MEDLINE/PubMed |
| BookMark | eNp9kUtPGzEUha2Kqjz_QIUq72AzqR_z8CxRCjRSCAuC6G7kca6D6Yyd2h5Q1D9fh6SoYlFvrnXvd2zdcw7RnnUWEPpMyYhSUn-dz2bTuxEjjI84K4pK8A_ogNGSZYwLsfd2r37so5MQnkg6JSnKvP6E9nlFOCkFOUC_J_3Ku2djl_gbwArPYPCySyW-OP8znOG5l8Zuxtp5POnlEvC4kyEYbZSMxln8YOIjnjnbGQvS47GzT8NSRsDXXi4M2JjdxXUH-GIhV9E8A75xfeoO_TH6qGUX4GRXj9D91eV8_D2b3l5PxhfTTPFSxKxmSmkKqio5B00pL4gURctyVciS5bJlbV4stGqlSqvnpWY6B1CK6KpuaaX5ETrfvps2_TVAiE1vgoKukxbcEBpW1aKivBY8oV926ND2sGhW3vTSr5u_hiWAbQHlXQge9BtCSbMJpnkNptkE0-yCSSLxTqRMfDUvJne7_0tPt1IDAP_8RQQlRc3_ADBznVI |
| CODEN | ITNNAL |
| CitedBy_id | crossref_primary_10_1109_TNNLS_2025_3548439 crossref_primary_10_1007_s11222_024_10409_5 crossref_primary_10_1109_TNNLS_2024_3395064 |
| Cites_doi | 10.1109/ICMEW.2017.8026312 10.1109/ICASSP.2013.6639349 10.1137/21m1453311 10.1137/16M1095056 10.1109/SP.2017.49 10.1137/0802003 10.1007/s10107-013-0677-5 10.1137/18M1224568 10.1007/978-3-319-46493-0_38 10.1137/1.9780898718003 10.1090/psapm/006/0084178 10.1109/ICCVW.2017.95 10.1007/BF01580369 10.1093/comjnl/7.2.149 10.1109/CDC.2014.7040179 10.1093/imanum/5.1.121 10.1016/0041-5553(85)90100-4 10.1109/CVPR.2016.90 10.1137/030601880 10.1007/s11263-015-0816-y 10.1051/m2an/196903R100351 10.1137/S1064827597323415 10.1090/S0025-5718-99-01171-0 10.5555/3454287.3455008 10.1016/S0893-6080(05)80056-5 10.1137/16M1080173 10.1038/nature14539 10.1007/s10208-013-9150-3 10.1016/0041-5553(64)90137-5 10.1137/s1052623497318992 |
| ContentType | Journal Article |
| DBID | 97E RIA RIE AAYXX CITATION NPM 7X8 |
| DOI | 10.1109/TNNLS.2023.3255783 |
| DatabaseName | IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef PubMed MEDLINE - Academic |
| DatabaseTitle | CrossRef PubMed MEDLINE - Academic |
| DatabaseTitleList | MEDLINE - Academic PubMed |
| Database_xml | – sequence: 1 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 2 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher – sequence: 3 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Computer Science |
| EISSN | 2162-2388 |
| EndPage | 12300 |
| ExternalDocumentID | 37030680 10_1109_TNNLS_2023_3255783 10081059 |
| Genre | orig-research Journal Article |
| GrantInformation_xml | – fundername: NSF grantid: DMS-1924935; DMS-1952339; DMS-2110145; DMS-2152762; DMS-2208361 funderid: 10.13039/100000001 – fundername: Office of Science of the Department of Energy grantid: DE-SC00211424; DE-SC0023490 funderid: 10.13039/100006132 – fundername: NSF grantid: DMS-1821144; DMS-2208314 funderid: 10.13039/100000001 |
| GroupedDBID | 0R~ 4.4 5VS 6IK 97E AAJGR AARMG AASAJ AAWTH ABAZT ABQJQ ABVLG ACIWK ACPRK AENEX AFRAH AGQYO AGSQL AHBIQ AKJIK AKQYR ALMA_UNASSIGNED_HOLDINGS ATWAV BEFXN BFFAM BGNUA BKEBE BPEOZ EBS EJD IFIPE IPLJI JAVBF M43 MS~ O9- OCL PQQKQ RIA RIE RNS AAYXX CITATION NPM RIG 7X8 |
| ID | FETCH-LOGICAL-c368t-92ccf1ec7633ef11350a85b24c5a624ab2b45dfcbac23846f2f4eecc0f79b17f3 |
| IEDL.DBID | RIE |
| ISICitedReferencesCount | 5 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000966203200001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 2162-237X 2162-2388 |
| IngestDate | Sun Sep 28 12:31:01 EDT 2025 Thu Jan 02 22:38:33 EST 2025 Tue Nov 18 22:13:15 EST 2025 Sat Nov 29 01:40:25 EST 2025 Wed Aug 27 02:02:08 EDT 2025 |
| IsDoiOpenAccess | false |
| IsOpenAccess | true |
| IsPeerReviewed | false |
| IsScholarly | true |
| Issue | 9 |
| Language | English |
| License | https://ieeexplore.ieee.org/Xplorehelp/downloads/license-information/IEEE.html https://doi.org/10.15223/policy-029 https://doi.org/10.15223/policy-037 |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c368t-92ccf1ec7633ef11350a85b24c5a624ab2b45dfcbac23846f2f4eecc0f79b17f3 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 |
| ORCID | 0000-0002-4848-4791 |
| OpenAccessLink | https://www.osti.gov/biblio/2280651 |
| PMID | 37030680 |
| PQID | 2798713983 |
| PQPubID | 23479 |
| PageCount | 13 |
| ParticipantIDs | ieee_primary_10081059 crossref_primary_10_1109_TNNLS_2023_3255783 pubmed_primary_37030680 crossref_citationtrail_10_1109_TNNLS_2023_3255783 proquest_miscellaneous_2798713983 |
| PublicationCentury | 2000 |
| PublicationDate | 2024-09-01 |
| PublicationDateYYYYMMDD | 2024-09-01 |
| PublicationDate_xml | – month: 09 year: 2024 text: 2024-09-01 day: 01 |
| PublicationDecade | 2020 |
| PublicationPlace | United States |
| PublicationPlace_xml | – name: United States |
| PublicationTitle | IEEE transaction on neural networks and learning systems |
| PublicationTitleAbbrev | TNNLS |
| PublicationTitleAlternate | IEEE Trans Neural Netw Learn Syst |
| PublicationYear | 2024 |
| Publisher | IEEE |
| Publisher_xml | – name: IEEE |
| References | ref13 ref12 ref56 ref15 Le (ref23) ref52 ref11 ref55 Duchi (ref9) 2011; 12 ref10 Sutskever (ref48) ref17 ref19 ref18 Zeiler (ref54) 2012 Dozat (ref8) Carmon (ref5) ref41 ref44 ref43 Zoutendijk (ref57) 1970 Goodfellow (ref14) 2014 ref49 Vapnik (ref50) Kingma (ref21) 2014 Nguyen (ref35); 190 ref7 ref4 Reddi (ref42) 2019 ref3 ref6 Nesterov (ref33) 1983; 269 ref40 Saad (ref45) 2003 Hinton (ref20) 2012; 14 Loshchilov (ref27) Liu (ref26) ref37 Xia (ref53); 34 Nguyen (ref34); 33 ref36 Liu (ref25) 2019 ref31 ref30 Wang (ref51) ref2 ref1 ref39 ref38 Krizhevsky (ref22) 2009 Nesterov (ref32) 1998; 3 Loshchilov (ref28) ref24 Su (ref47) Shewchuck (ref46) 1994 Hardt (ref16) 2014 Madry (ref29) 2017 |
| References_xml | – ident: ref36 doi: 10.1109/ICMEW.2017.8026312 – volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref51 article-title: ResNet ensemble via the Feynman–Kac formalism to improve natural and robust acurcies – start-page: 10705 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref5 article-title: Analysis of Krylov subspace solutions of regularized non-convex quadratic problems – volume: 190 start-page: 189 volume-title: Proc. Math. Sci. Mach. Learn. ident: ref35 article-title: Momentum transformer: Closing the performance gap between self-attention and its linearization – ident: ref2 doi: 10.1109/ICASSP.2013.6639349 – year: 2014 ident: ref21 article-title: Adam: A method for stochastic optimization publication-title: arXiv:1412.6980 – ident: ref52 doi: 10.1137/21m1453311 – ident: ref55 doi: 10.1137/16M1095056 – ident: ref4 doi: 10.1109/SP.2017.49 – ident: ref11 doi: 10.1137/0802003 – ident: ref7 doi: 10.1007/s10107-013-0677-5 – ident: ref43 doi: 10.1137/18M1224568 – ident: ref18 doi: 10.1007/978-3-319-46493-0_38 – volume-title: Iterative Methods for Sparse Linear Systems year: 2003 ident: ref45 doi: 10.1137/1.9780898718003 – start-page: 1139 volume-title: Proc. Int. Conf. Mach. Learn. ident: ref48 article-title: On the importance of initialization and momentum in deep learning – ident: ref19 doi: 10.1090/psapm/006/0084178 – ident: ref56 doi: 10.1109/ICCVW.2017.95 – start-page: 37 volume-title: Integer and Nonlinear Programming year: 1970 ident: ref57 article-title: Nonlinear programming, computational methods – volume: 12 start-page: 2121 year: 2011 ident: ref9 article-title: Adaptive subgradient methods for online learning and stochastic optimization publication-title: J. Mach. Learn. Res. – ident: ref41 doi: 10.1007/BF01580369 – volume: 3 start-page: 5 issue: 4 year: 1998 ident: ref32 article-title: Introductory lectures on convex programming volume I: Basic course publication-title: Lecture Notes – start-page: 265 volume-title: Proc. 28th Int. Conf. Int. Conf. Mach. Learn. ident: ref23 article-title: On optimization methods for deep learning – ident: ref10 doi: 10.1093/comjnl/7.2.149 – ident: ref12 doi: 10.1109/CDC.2014.7040179 – year: 2014 ident: ref14 article-title: Explaining and harnessing adversarial examples publication-title: arXiv:1412.6572 – ident: ref1 doi: 10.1093/imanum/5.1.121 – volume: 33 start-page: 1924 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref34 article-title: Momentumrnn: Integrating momentum into recurrent neural networks – year: 2017 ident: ref29 article-title: Towards deep learning models resistant to adversarial attacks publication-title: arXiv:1706.06083 – ident: ref31 doi: 10.1016/0041-5553(85)90100-4 – ident: ref17 doi: 10.1109/CVPR.2016.90 – start-page: 1 volume-title: Proc. ICLR ident: ref28 article-title: Fixing weight decay regularization in Adam – ident: ref15 doi: 10.1137/030601880 – year: 1994 ident: ref46 article-title: An introduction to the conjugate gradient method without agonizing pain – year: 2012 ident: ref54 article-title: ADADELTA: An adaptive learning rate method publication-title: arXiv:1212.5701 – volume: 14 start-page: 2 issue: 8 year: 2012 ident: ref20 article-title: Neural networks for machine learning lecture 6A overview of mini-batch gradient descent – start-page: 2510 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref47 article-title: A differential equation for modeling Nesterov’s accelerated gradient method: Theory and insights – year: 2019 ident: ref42 article-title: On the convergence of Adam and beyond publication-title: arXiv:1904.09237 – volume-title: Robustness Versus Acceleration year: 2014 ident: ref16 – ident: ref44 doi: 10.1007/s11263-015-0816-y – ident: ref39 doi: 10.1051/m2an/196903R100351 – volume-title: Radam year: 2019 ident: ref25 – ident: ref13 doi: 10.1137/S1064827597323415 – ident: ref49 doi: 10.1090/S0025-5718-99-01171-0 – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref26 article-title: On the variance of the adaptive learning rate and beyond – volume: 269 start-page: 543 issue: 3 year: 1983 ident: ref33 article-title: A method for solving the convex programming problem with convergence rate O(1/k2) publication-title: Doklady Akademii Nauk SSSR – ident: ref38 doi: 10.5555/3454287.3455008 – volume: 34 start-page: 18646 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref53 article-title: Heavy ball neural ordinary differential equations – start-page: 831 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref50 article-title: Principles of risk minimization for learning theory – ident: ref30 doi: 10.1016/S0893-6080(05)80056-5 – start-page: 1 volume-title: Proc. ICLR ident: ref8 article-title: Incorporating Nesterov momentum into Adam – ident: ref3 doi: 10.1137/16M1080173 – volume-title: Learning multiple layers of features from tiny images year: 2009 ident: ref22 – ident: ref24 doi: 10.1038/nature14539 – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref27 article-title: Decoupled weight decay regularization – ident: ref37 doi: 10.1007/s10208-013-9150-3 – ident: ref40 doi: 10.1016/0041-5553(64)90137-5 – ident: ref6 doi: 10.1137/s1052623497318992 |
| SSID | ssj0000605649 |
| Score | 2.483582 |
| Snippet | Momentum is crucial in stochastic gradient-based optimization algorithms for accelerating or improving training deep neural networks (DNNs). In deep learning... |
| SourceID | proquest pubmed crossref ieee |
| SourceType | Aggregation Database Index Database Enrichment Source Publisher |
| StartPage | 12288 |
| SubjectTerms | Adaptive momentum Classification algorithms Convergence Deep learning Image classification nonlinear conjugate gradient (NCG) Robustness Stochastic processes Training |
| Title | Improving Deep Neural Networks' Training for Image Classification With Nonlinear Conjugate Gradient-Style Adaptive Momentum |
| URI | https://ieeexplore.ieee.org/document/10081059 https://www.ncbi.nlm.nih.gov/pubmed/37030680 https://www.proquest.com/docview/2798713983 |
| Volume | 35 |
| WOSCitedRecordID | wos000966203200001&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: PRVIEE databaseName: IEEE Electronic Library (IEL) customDbUrl: eissn: 2162-2388 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000605649 issn: 2162-237X databaseCode: RIE dateStart: 20120101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3BatwwEBXN0kMvTdomzSZtUKGQQ_FGlmTLPi5p0hZSU8iG7M1IskRTEu-yaxdKfj4zsr3kkkBvxnhswxtJb6SZeYR8diyvmPYsynlcRTIzJjLGpxHerYxULHNBteRCFUU2n-e_-mL1UAvjnAvJZ26Cl-Esv1rYFrfKTrARDfKBLbKlVNoVa202VBgQ8zTQXR6nPOJCzYciGZafzIri4nKCWuETASxaZSigI9DdU-wI-WhNCiIrT_PNsO6cb__nH--Q1z3BpNPOI96QF65-S7YH8Qbaj-V35H6znUC_Orek2KUD7IouLXx9TGe9eAQFWkt_3MG8Q4OCJuYWBTjp9U3zmxZdrw29oqeL-k-Lu3L02yokkjXRZfPv1tFppZc4q9Kf2O6hae92ydX52ez0e9QrMURWpFkDQFrrY2dhMhLOx7FImM4Sw6VNdMqlNtzIpPLWaAsUQKaee-nAOZhXuYmVF3tkVC9qt09o5Zk0GniTVAJCQzD0JjYQWImsSoz1YxIPWJS2b1OOahm3ZQhXWF4GKEuEsuyhHJMvG5tl16Tj2ad3EahHT3YYjcmnAfMSxhgenOjaLdp1yVUOcaXI0fZ95wwb68GHDp546yF5BR-XXVraBzJqVq37SF7av83NenUEjjzPjoIjPwB1DO6h |
| linkProvider | IEEE |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1La9wwEB6aNNBcmj6SdJukVSHQQ_GuLMmvY8ijCd2YQDZ0b0aSJZqSeJddu1D656OR7SWXFHIzxmME30j6RpqZD-DQ0Kyk0tIgY2EZiFSpQCkbB_i2VCKhqfGqJeMkz9PpNLvqitV9LYwxxiefmSE--rv8cqYbPCobYSMa5ANr8DISgtG2XGt1pEIdNY894WVhzALGk2lfJkOz0STPx9dDVAsfcsejkxQldDg6fIw9IR_tSl5m5WnG6Xees61njvkNvO4oJjlqfeItvDDVO9jq5RtIN5vfw7_VgQI5MWZOsE-Hs8vbxPDlVzLp5COII7bk4t6tPMRraGJ2kQeU_Lytf5G87bYhF-R4Vv1u8FyOfF_4VLI6uK7_3hlyVMo5rqvkEhs-1M39NtycnU6Oz4NOiyHQPE5rB6XWNjTaLUfc2DDkEZVppJjQkYyZkIopEZVWK6kdCRCxZVYY5x7UJpkKE8t3YL2aVeYDkNJSoaRjTiLhLjh0hlaFyoVWPC0jpe0Awh6LQneNylEv467wAQvNCg9lgVAWHZQD-LaymbdtOv779TYC9ejLFqMBfOkxL9wsw6sTWZlZsyxYkrnIkmdou9s6w8q696GPT_z1M7w6n1yOi_FF_mMPNt1ARJuktg_r9aIxB7Ch_9S3y8Un784PYlPxAA |
| 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=Improving+Deep+Neural+Networks%27+Training+for+Image+Classification+With+Nonlinear+Conjugate+Gradient-Style+Adaptive+Momentum&rft.jtitle=IEEE+transaction+on+neural+networks+and+learning+systems&rft.au=Wang%2C+Bao&rft.au=Ye%2C+Qiang&rft.date=2024-09-01&rft.issn=2162-2388&rft.eissn=2162-2388&rft.volume=35&rft.issue=9&rft.spage=12288&rft_id=info:doi/10.1109%2FTNNLS.2023.3255783&rft.externalDBID=NO_FULL_TEXT |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2162-237X&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2162-237X&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2162-237X&client=summon |