TCL: Time-Dependent Clustering Loss for Optimizing Post-Training Feature Map Quantization for Partitioned DNNs

This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning, autoencoder-based compression, quantization with Time Dependent Clustering Loss (TCL) regularization, and lossless compression, to reduce communication o...

Celý popis

Uložené v:

Podrobná bibliografia
Vydané v:	IEEE access Ročník 13; s. 103640 - 103648
Hlavní autori:	Berg, Oscar Artur Bernd, Saqib, Eiraj, Jantsch, Axel, Shallari, Irida, Krug, Silvia, Sanchez Leal, Isaac, O'Nils, Mattias
Médium:	Journal Article
Jazyk:	English
Vydavateľské údaje:	Piscataway IEEE 2025 The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Predmet:	Accuracy Adaptation models Autoencoders Clustering CNN Computational modeling Constraints Feature maps Internet of Things IoT Load modeling Machine learning Object detection Partitioning Quantization Quantization (signal) Real time Regularization Servers Time dependence Training
ISSN:	2169-3536, 2169-3536
On-line prístup:	Získať plný text
Tagy:	Pridať tag Žiadne tagy, Buďte prvý, kto otaguje tento záznam!

Abstract	This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning, autoencoder-based compression, quantization with Time Dependent Clustering Loss (TCL) regularization, and lossless compression, to reduce communication overhead, minimizing latency while maintaining accuracy. The autoencoder compresses feature maps at the partitioning point before quantization, effectively reducing data size and preserving accuracy. TCL regularization clusters activations at the partitioning point to align with quantization levels, minimizing quantization error and ensuring accuracy even with extreme low-bitwidth quantization. Our method is evaluated on classification models (ResNet-50, EfficientNetV2-S) and an object detection model (YOLOv10n) using the TinyImageNet-200 and Pascal VOC datasets. Deployed on Raspberry Pi 4 B and GPU, each model is tested across various partitioning points, quantization bit-widths (1-bit, 2-bit, and 3-bit), communication datarate (1MB/s to 10MB/s), and LZMA lossless compression. For a partitioned ResNet-50 after the convolutional stem block, the speed-up against a server solution is <inline-formula> <tex-math notation="LaTeX">2.33\times </tex-math></inline-formula> and 1.85x compared to the all-in-node solution, with only a minimal accuracy drop of less than one percentage points. The proposed framework offers a scalable solution for deploying high-performance AI models on IoT devices, extending the feasibility of real-time inference in resource-constrained environments.
AbstractList	This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning, autoencoder-based compression, quantization with Time Dependent Clustering Loss (TCL) regularization, and lossless compression, to reduce communication overhead, minimizing latency while maintaining accuracy. The autoencoder compresses feature maps at the partitioning point before quantization, effectively reducing data size and preserving accuracy. TCL regularization clusters activations at the partitioning point to align with quantization levels, minimizing quantization error and ensuring accuracy even with extreme low-bitwidth quantization. Our method is evaluated on classification models (ResNet-50, EfficientNetV2-S) and an object detection model (YOLOv10n) using the TinyImageNet-200 and Pascal VOC datasets. Deployed on Raspberry Pi 4 B and GPU, each model is tested across various partitioning points, quantization bit-widths (1-bit, 2-bit, and 3-bit), communication datarate (1MB/s to 10MB/s), and LZMA lossless compression. For a partitioned ResNet-50 after the convolutional stem block, the speed-up against a server solution is [Formula Omitted] and 1.85x compared to the all-in-node solution, with only a minimal accuracy drop of less than one percentage points. The proposed framework offers a scalable solution for deploying high-performance AI models on IoT devices, extending the feasibility of real-time inference in resource-constrained environments. This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning, autoencoder-based compression, quantization with Time Dependent Clustering Loss (TCL) regularization, and lossless compression, to reduce communication overhead, minimizing latency while maintaining accuracy. The autoencoder compresses feature maps at the partitioning point before quantization, effectively reducing data size and preserving accuracy. TCL regularization clusters activations at the partitioning point to align with quantization levels, minimizing quantization error and ensuring accuracy even with extreme low-bitwidth quantization. Our method is evaluated on classification models (ResNet-50, EfficientNetV2-S) and an object detection model (YOLOv10n) using the TinyImageNet-200 and Pascal VOC datasets. Deployed on Raspberry Pi 4 B and GPU, each model is tested across various partitioning points, quantization bit-widths (1-bit, 2-bit, and 3-bit), communication datarate (1MB/s to 10MB/s), and LZMA lossless compression. For a partitioned ResNet-50 after the convolutional stem block, the speed-up against a server solution is <inline-formula> <tex-math notation="LaTeX">2.33\times </tex-math></inline-formula> and 1.85x compared to the all-in-node solution, with only a minimal accuracy drop of less than one percentage points. The proposed framework offers a scalable solution for deploying high-performance AI models on IoT devices, extending the feasibility of real-time inference in resource-constrained environments. This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning, autoencoder-based compression, quantization with Time Dependent Clustering Loss (TCL) regularization, and lossless compression, to reduce communication overhead, minimizing latency while maintaining accuracy. The autoencoder compresses feature maps at the partitioning point before quantization, effectively reducing data size and preserving accuracy. TCL regularization clusters activations at the partitioning point to align with quantization levels, minimizing quantization error and ensuring accuracy even with extreme low-bitwidth quantization. Our method is evaluated on classification models (ResNet-50, EfficientNetV2-S) and an object detection model (YOLOv10n) using the TinyImageNet-200 and Pascal VOC datasets. Deployed on Raspberry Pi 4 B and GPU, each model is tested across various partitioning points, quantization bit-widths (1-bit, 2-bit, and 3-bit), communication datarate (1MB/s to 10MB/s), and LZMA lossless compression. For a partitioned ResNet-50 after the convolutional stem block, the speed-up against a server solution is <tex-math notation="LaTeX">$2.33\times $ </tex-math> and 1.85x compared to the all-in-node solution, with only a minimal accuracy drop of less than one percentage points. The proposed framework offers a scalable solution for deploying high-performance AI models on IoT devices, extending the feasibility of real-time inference in resource-constrained environments. This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning, autoencoder-based compression, quantization with Time Dependent Clustering Loss (TCL) regularization, and lossless compression, to reduce communication overhead, minimizing latency while maintaining accuracy. The autoencoder compresses feature maps at the partitioning point before quantization, effectively reducing data size and preserving accuracy. TCL regularization clusters activations at the partitioning point to align with quantization levels, minimizing quantization error and ensuring accuracy even with extreme low-bitwidth quantization. Our method is evaluated on classification models (ResNet-50, EfficientNetV2-S) and an object detection model (YOLOv10n) using the TinyImageNet-200 and Pascal VOC datasets. Deployed on Raspberry Pi 4 B and GPU, each model is tested across various partitioning points, quantization bit-widths (1-bit, 2-bit, and 3-bit), communication datarate (1MB/s to 10MB/s), and LZMA lossless compression. For a partitioned ResNet-50 after the convolutional stem block, the speed-up against a server solution is 2.33× and 1.85x compared to the all-in-node solution, with only a minimal accuracy drop of less than one percentage points. The proposed framework offers a scalable solution for deploying high-performance AI models on IoT devices, extending the feasibility of real-time inference in resource-constrained environments.
Author	Jantsch, Axel O'Nils, Mattias Krug, Silvia Berg, Oscar Artur Bernd Saqib, Eiraj Sanchez Leal, Isaac Shallari, Irida
Author_xml	– sequence: 1 givenname: Oscar Artur Bernd orcidid: 0009-0000-8343-9649 surname: Berg fullname: Berg, Oscar Artur Bernd organization: Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden – sequence: 2 givenname: Eiraj orcidid: 0000-0002-9903-1338 surname: Saqib fullname: Saqib, Eiraj organization: Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden – sequence: 3 givenname: Axel orcidid: 0000-0003-2251-0004 surname: Jantsch fullname: Jantsch, Axel organization: Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden – sequence: 4 givenname: Irida orcidid: 0000-0002-3774-4850 surname: Shallari fullname: Shallari, Irida organization: Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden – sequence: 5 givenname: Silvia orcidid: 0000-0003-0282-5471 surname: Krug fullname: Krug, Silvia organization: Institut für Mikroelektronik- und Mechatronik-Systeme gemeinnützige GmbH (IMMS GmbH), Ilmenau, Germany – sequence: 6 givenname: Isaac orcidid: 0000-0002-3351-0491 surname: Sanchez Leal fullname: Sanchez Leal, Isaac organization: Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden – sequence: 7 givenname: Mattias orcidid: 0000-0001-8607-4083 surname: O'Nils fullname: O'Nils, Mattias email: mattias.onils@miun.se organization: Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden
BackLink	https://urn.kb.se/resolve?urn=urn:nbn:se:miun:diva-54756$$DView record from Swedish Publication Index (Mittuniversitetet)
BookMark	eNpVkVtv1DAQhSNUJErpL4CHSDxn8SW2Y95W2RYqLW1RF14tO55UXu3awXaE6K8n2VRc_GLP6JxPMz6vizMfPBTFW4xWGCP5Yd22Vw8PK4IIW1EmJEbiRXFOMJcVZZSf_fN-VVymtEfTaaYWE-eF37Xbj-XOHaHawADegs9lexhThuj8Y7kNKZV9iOXdkN3RPc29-5BytYva-bm6Bp3HCOUXPZRfR-2ze9LZBX9y3euY3VyBLTe3t-lN8bLXhwSXz_dF8e36atd-rrZ3n27a9bbqqJS5skZ0xkjaaCa5kTXQusNYyqZGmCPDLZbACaq11ESYhnPeadb3XAgpuJWWXhQ3C9cGvVdDdEcdf6mgnTo1QnxU82TdARQYbBvKiLFM1rqmujHGmqYBYYFY1E-samGlnzCM5j_axn1fn2hHN3rFasH4pH-_6IcYfoyQstqHMfppXUUJIVxMm7BJRRdVF6cvjtD_4WKk5lzVkquac1XPuU6ud4vLAcBfB0YU10zQ3_r4oP4
CODEN	IAECCG
Cites_doi	10.1609/aaai.v35i3.16351 10.1145/3384419.3430898 10.1109/SENSORS56945.2023.10324987 10.1109/ACCESS.2019.2913703 10.1109/SAS58821.2023.10254054 10.1109/COMSNETS48256.2020.9027432 10.1109/icc45041.2023.10279110 10.1109/COMST.2024.3393230 10.1109/COMST.2022.3218527 10.1109/JPROC.2020.2976475 10.1145/3576914.3587518 10.1109/MNET.2024.3420755 10.1186/s13677-023-00493-9 10.1109/ICME.2018.8486500 10.1145/3093336.3037698 10.1109/ISLPED.2019.8824955 10.3390/computers12030060
ContentType	Journal Article
Copyright	Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2025
Copyright_xml	– notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2025
DBID	97E ESBDL RIA RIE AAYXX CITATION 7SC 7SP 7SR 8BQ 8FD JG9 JQ2 L7M L~C L~D ADTPV AKRZP AOWAS D8T DG5 ZZAVC DOA
DOI	10.1109/ACCESS.2025.3579107
DatabaseName	IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE Xplore Open Access Journals IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE Electronic Library (IEL) CrossRef Computer and Information Systems Abstracts Electronics & Communications Abstracts Engineered Materials Abstracts METADEX Technology Research Database Materials Research Database ProQuest Computer Science Collection Advanced Technologies Database with Aerospace Computer and Information Systems Abstracts Academic Computer and Information Systems Abstracts Professional SwePub SWEPUB Mittuniversitetet full text SwePub Articles SWEPUB Freely available online SWEPUB Mittuniversitetet SwePub Articles full text DOAJ Directory of Open Access Journals
DatabaseTitle	CrossRef Materials Research Database Engineered Materials Abstracts Technology Research Database Computer and Information Systems Abstracts – Academic Electronics & Communications Abstracts ProQuest Computer Science Collection Computer and Information Systems Abstracts Advanced Technologies Database with Aerospace METADEX Computer and Information Systems Abstracts Professional
DatabaseTitleList	Materials Research Database
Database_xml	– sequence: 1 dbid: DOA name: DOAJ Directory of Open Access Journals url: https://www.doaj.org/ sourceTypes: Open Website – sequence: 2 dbid: RIE name: IEEE Xplore url: https://ieeexplore.ieee.org/ sourceTypes: Publisher
DeliveryMethod	fulltext_linktorsrc
Discipline	Engineering
EISSN	2169-3536
EndPage	103648
ExternalDocumentID	oai_doaj_org_article_eb1d8352bd594a43a8bbdb88e7de2d0f oai_DiVA_org_miun_54756 10_1109_ACCESS_2025_3579107 11031457
Genre	orig-research
GrantInformation_xml	– fundername: The Swedish Knowledge Foundation funderid: 10.13039/100003077
GroupedDBID	0R~ 4.4 5VS 6IK 97E AAJGR ABAZT ABVLG ACGFS ADBBV AGSQL ALMA_UNASSIGNED_HOLDINGS BCNDV BEFXN BFFAM BGNUA BKEBE BPEOZ EBS EJD ESBDL GROUPED_DOAJ IPLJI JAVBF KQ8 M43 M~E O9- OCL OK1 RIA RIE RNS AAYXX CITATION 7SC 7SP 7SR 8BQ 8FD JG9 JQ2 L7M L~C L~D ADTPV AKRZP AOWAS D8T DG5 ZZAVC
ID	FETCH-LOGICAL-c399t-db7cbb938a596b94e34c1199840160b6d19e6204a9a27b8666ca5ff677976d9d3
IEDL.DBID	RIE
ISICitedReferencesCount	0
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001512606800010&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN	2169-3536
IngestDate	Fri Oct 03 12:51:13 EDT 2025 Tue Nov 04 17:02:34 EST 2025 Mon Dec 08 03:37:32 EST 2025 Sat Nov 29 07:49:55 EST 2025 Wed Aug 27 01:46:20 EDT 2025
IsDoiOpenAccess	true
IsOpenAccess	true
IsPeerReviewed	true
IsScholarly	true
Language	English
License	https://creativecommons.org/licenses/by/4.0/legalcode
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-c399t-db7cbb938a596b94e34c1199840160b6d19e6204a9a27b8666ca5ff677976d9d3
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ORCID	0000-0001-8607-4083 0000-0002-9903-1338 0009-0000-8343-9649 0000-0003-0282-5471 0000-0002-3351-0491 0000-0003-2251-0004 0000-0002-3774-4850
OpenAccessLink	https://ieeexplore.ieee.org/document/11031457
PQID	3222671195
PQPubID	4845423
PageCount	9
ParticipantIDs	doaj_primary_oai_doaj_org_article_eb1d8352bd594a43a8bbdb88e7de2d0f ieee_primary_11031457 crossref_primary_10_1109_ACCESS_2025_3579107 proquest_journals_3222671195 swepub_primary_oai_DiVA_org_miun_54756
PublicationCentury	2000
PublicationDate	20250000 2025-00-00 20250101 2025 2025-01-01
PublicationDateYYYYMMDD	2025-01-01
PublicationDate_xml	– year: 2025 text: 20250000
PublicationDecade	2020
PublicationPlace	Piscataway
PublicationPlace_xml	– name: Piscataway
PublicationTitle	IEEE access
PublicationTitleAbbrev	Access
PublicationYear	2025
Publisher	IEEE The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Publisher_xml	– name: IEEE – name: The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
References	ref13 ref12 ref23 ref14 ref20 ref11 ref22 ref10 Polino (ref19) 2018 ref21 Hubara (ref15) 2020 ref2 ref1 Mishra (ref4) 2020 Darabi (ref17) ref18 Sung (ref5) 2015 ref8 ref7 ref9 Esser (ref16) 2019 ref6 Li (ref3) 2023; 12
References_xml	– ident: ref7 doi: 10.1609/aaai.v35i3.16351 – year: 2020 ident: ref15 article-title: Improving post training neural quantization: Layer-wise calibration and integer programming publication-title: arXiv:2006.10518 – ident: ref22 doi: 10.1145/3384419.3430898 – ident: ref2 doi: 10.1109/SENSORS56945.2023.10324987 – ident: ref23 doi: 10.1109/ACCESS.2019.2913703 – ident: ref8 doi: 10.1109/SAS58821.2023.10254054 – ident: ref12 doi: 10.1109/COMSNETS48256.2020.9027432 – year: 2015 ident: ref5 article-title: Resiliency of deep neural networks under quantization publication-title: arXiv:1511.06488 – year: 2018 ident: ref19 article-title: Model compression via distillation and quantization publication-title: arXiv:1802.05668 – ident: ref21 doi: 10.1109/icc45041.2023.10279110 – ident: ref9 doi: 10.1109/COMST.2024.3393230 – year: 2020 ident: ref4 article-title: A survey on deep neural network compression: Challenges, overview, and solutions publication-title: arXiv:2010.03954 – start-page: 1 volume-title: Proc. ICLR ident: ref17 article-title: BNN+: Improved binary network training – year: 2019 ident: ref16 article-title: Learned step size quantization publication-title: arXiv:1902.08153 – ident: ref10 doi: 10.1109/COMST.2022.3218527 – ident: ref1 doi: 10.1109/JPROC.2020.2976475 – ident: ref6 doi: 10.1145/3576914.3587518 – ident: ref14 doi: 10.1109/MNET.2024.3420755 – ident: ref13 doi: 10.1186/s13677-023-00493-9 – ident: ref18 doi: 10.1109/ICME.2018.8486500 – ident: ref11 doi: 10.1145/3093336.3037698 – ident: ref20 doi: 10.1109/ISLPED.2019.8824955 – volume: 12 start-page: 60 issue: 3 year: 2023 ident: ref3 article-title: Model compression for deep neural networks: A survey publication-title: Computers doi: 10.3390/computers12030060
SSID	ssj0000816957
Score	2.3342764
Snippet	This paper introduces an enhanced approach for deploying deep learning models on resource-constrained IoT devices by combining model partitioning,...
SourceID	doaj swepub proquest crossref ieee
SourceType	Open Website Open Access Repository Aggregation Database Index Database Publisher
StartPage	103640
SubjectTerms	Accuracy Adaptation models Autoencoders Clustering CNN Computational modeling Constraints Feature maps Internet of Things IoT Load modeling Machine learning Object detection Partitioning Quantization Quantization (signal) Real time Regularization Servers Time dependence Training
SummonAdditionalLinks	– databaseName: DOAJ Directory of Open Access Journals dbid: DOA link: http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwrV3PT9swFLYmtAM7TBtjWjdAPkw7zSNp7NjmVlIQh65jUoe4Wf4VqRJUqG124K_nPcdAe-LCNXLy7O8lz99L8r5HyHevhapKbllRW8s4cGbmStsy553UroANOdWtXU3kdKqur_XlRqsv_CeslwfugTuGWBKQJbggNLe8ssq54JSKMsRhKFqMvoXUG8lUisGqrLWQWWaoLPTxqGlgRZAQDsWvSkiYgtzaipJif26xss02NxVE065z_oG8z3SRjvppfiRv4mKPvNsQEfxEFrNmckKxlIONc0fbNW1uOlRAgAF0ArYpUFP6B6LD7fwej2GLXjbL3SEo0sBuGelve0f_dgB1rs1MZ10iREnPKNDxdLraJ__Oz2bNBctdFJgH8rFmwUnvnK6UFbp2mseK-xIr6ziKy7k6lDqiKL3VdiidgnTGW9G2tZTAVIIO1WeyswAjXwiF1Mp74Yuq9YoDThrzSVHJItZRWmcH5OcjoOauF8swKckotOnxN4i_yfgPyCmC_jQUla7TAfC_yf43L_l_QPbRZc_2sG8FF3Dxg0cfmvxYrgx-VqolqtwNyI_er1vWx_OrUbJ-O-8WRnAp6q-vMclvZBcX3r-6OSA762UXD8lb_389Xy2P0q37AFno8u8 priority: 102 providerName: Directory of Open Access Journals
Title	TCL: Time-Dependent Clustering Loss for Optimizing Post-Training Feature Map Quantization for Partitioned DNNs
URI	https://ieeexplore.ieee.org/document/11031457 https://www.proquest.com/docview/3222671195 https://urn.kb.se/resolve?urn=urn:nbn:se:miun:diva-54756 https://doaj.org/article/eb1d8352bd594a43a8bbdb88e7de2d0f
Volume	13
WOSCitedRecordID	wos001512606800010&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: PRVAON databaseName: DOAJ Directory of Open Access Journals customDbUrl: eissn: 2169-3536 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000816957 issn: 2169-3536 databaseCode: DOA dateStart: 20130101 isFulltext: true titleUrlDefault: https://www.doaj.org/ providerName: Directory of Open Access Journals – providerCode: PRVHPJ databaseName: ROAD: Directory of Open Access Scholarly Resources customDbUrl: eissn: 2169-3536 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0000816957 issn: 2169-3536 databaseCode: M~E dateStart: 20130101 isFulltext: true titleUrlDefault: https://road.issn.org providerName: ISSN International Centre
link	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Lb9QwEB7RigMceBaRUiofECfSJhs7jrkt2VYctkuRlqo3y69IK7VLtbvh0AO_nRnHLd0DBy5RZDkZ2984nrEz3wB8cEo0VclNXtTG5Bxt5tyWpsuts1LZAhfkGLd2MZWzWXN5qc5TsHqMhQkhxJ_PwhHdxrN8_9P1tFV2XFJOAi7kDuxIWQ_BWvcbKpRBQgmZmIXKQh2P2xY7gT7gSBxVQqJUubX6RJL-lFVl28B8SBoaF5rT5__ZxBfwLFmUbDyowEt4FJav4OkDnsHXsJy308-Moj3ySUp6u2HtVU8kCViBTbGtDK1X9g0_INeLWyqjLL75PCWQYGQp9qvAzswN-94jGil8Mz51TuoXKY88m8xm6z34cXoyb7_mKdFC7tA-2eTeSmetqhojVG0VDxV3JQXfceKfs7UvVSDeeqPMSNoGPR5nRNfVUqIx45Wv3sDuEoW8BYbel3PCFVXnGo6jo8jlFJUsQh2ksSaDT3cA6JuBT0NHP6RQesBLE1464ZXBFwLpviqRYccCHHSd5pbG5caTIWm9UNzwyjTWets0Qfow8kWXwR4B9VdewiiDgzvMdZq5a00nT7UkIrwMPg56sCV9srgYR-nXi36pBZei3v_H-9_BE-rLsGFzALubVR_ew2P3a7NYrw6j94_Xs98nh1GT_wCX6_EG
linkProvider	IEEE
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1Lb9QwEB5BQQIOPIsIFPABccJtsrHjmNuSpSoiDUVaqt4svyKt1C7V7oYDvx6P45bugQO3yHIysb9xZsbOfAPwzkpelwXTNK-0piz4zNQUuqfGGiFNHgxyzFs7bUXX1Wdn8iQlq8dcGO99_PnM7-NlPMt3P-2AW2UHBdYkYFzchjs8GNJqTNe63lLBGhKSi8QtVOTyYNo0YRghCpzw_ZKLIFds2Z9I05_qqmy7mDdpQ6OpOXz0ny_5GB4mn5JMRyV4Arf88ik8uME0-AyW86b9SDDfg85S2dsNac4HpEkIHUgb3pUE_5V8C5-Qi8VvbMM6vnSeSkgQ9BWHlSfH-pJ8HwIeKYEz3nWCChhJjxyZdd16F34cfp43RzSVWqA2eCgb6oywxsiy1lxWRjJfMltg-h1DBjpTuUJ6ZK7XUk-EqUPMYzXv-0qI4M446crnsLMMQl4ACfGXtdzmZW9rFmZHYtDJS5H7ygttdAYfrgBQlyOjhoqRSC7ViJdCvFTCK4NPCNJ1V6TDjg1h0lVaXSoYHIeupHFcMs1KXRvjTF174fzE5X0GuwjUX3kJowz2rjBXae2uFZ49VQKp8DJ4P-rBlvTZ4nQapV8shqXiTPDq5T-e_xbuHc2PW9V-6b6-gvs4rnH7Zg92NqvBv4a79tdmsV69iZr8Bwcn8jM
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=TCL%3A+Time-Dependent+Clustering+Loss+for+Optimizing+Post-Training+Feature+Map+Quantization+for+Partitioned+DNNs&rft.jtitle=IEEE+access&rft.au=Berg%2C+Oscar+Artur+Bernd&rft.au=Saqib%2C+Eiraj&rft.au=Jantsch%2C+Axel&rft.au=Shallari%2C+Irida&rft.date=2025&rft.pub=IEEE&rft.eissn=2169-3536&rft.volume=13&rft.spage=103640&rft.epage=103648&rft_id=info:doi/10.1109%2FACCESS.2025.3579107&rft.externalDocID=11031457
thumbnail_l	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=2169-3536&client=summon
thumbnail_m	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=2169-3536&client=summon
thumbnail_s	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=2169-3536&client=summon