Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review

Successful integration of deep neural networks (DNNs) or deep learning (DL) has resulted in breakthroughs in many areas. However, deploying these highly accurate models for data-driven, learned, automatic, and practical machine learning (ML) solutions to end-user applications remains challenging. DL...

Ausführliche Beschreibung

Gespeichert in:

Bibliographische Detailangaben
Veröffentlicht in:	Proceedings of the IEEE Jg. 111; H. 1; S. 1 - 50
Hauptverfasser:	Shuvo, Md. Maruf Hossain, Islam, Syed Kamrul, Cheng, Jianlin, Morshed, Bashir I.
Format:	Journal Article
Sprache:	Englisch
Veröffentlicht:	New York IEEE 01.01.2023 The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:	Algorithms Algorithm–hardware codesign Artificial intelligence artificial intelligence (AI) artificial intelligence on edge (edge-AI) Artificial neural networks Cloud computing Co-design Computer architecture Cyber-physical systems Data transmission Deep learning deep learning (DL) Design optimization Hardware Image edge detection Inference Internet of Things Iterative methods Machine learning Memory devices model compression Network latency neural accelerator Optimization Optimization techniques Real-time systems State-of-the-art reviews Training
ISSN:	0018-9219, 1558-2256
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

Abstract	Successful integration of deep neural networks (DNNs) or deep learning (DL) has resulted in breakthroughs in many areas. However, deploying these highly accurate models for data-driven, learned, automatic, and practical machine learning (ML) solutions to end-user applications remains challenging. DL algorithms are often computationally expensive, power-hungry, and require large memory to process complex and iterative operations of millions of parameters. Hence, training and inference of DL models are typically performed on high-performance computing (HPC) clusters in the cloud. Data transmission to the cloud results in high latency, round-trip delay, security and privacy concerns, and the inability of real-time decisions. Thus, processing on edge devices can significantly reduce cloud transmission cost. Edge devices are end devices closest to the user, such as mobile phones, cyber-physical systems (CPSs), wearables, the Internet of Things (IoT), embedded and autonomous systems, and intelligent sensors. These devices have limited memory, computing resources, and power-handling capability. Therefore, optimization techniques at both the hardware and software levels have been developed to handle the DL deployment efficiently on the edge. Understanding the existing research, challenges, and opportunities is fundamental to leveraging the next generation of edge devices with artificial intelligence (AI) capability. Mainly, four research directions have been pursued for efficient DL inference on edge devices: 1) novel DL architecture and algorithm design; 2) optimization of existing DL methods; 3) development of algorithm-hardware codesign; and 4) efficient accelerator design for DL deployment. This article focuses on surveying each of the four research directions, providing a comprehensive review of the state-of-the-art tools and techniques for efficient edge inference.
AbstractList	Successful integration of deep neural networks (DNNs) or deep learning (DL) has resulted in breakthroughs in many areas. However, deploying these highly accurate models for data-driven, learned, automatic, and practical machine learning (ML) solutions to end-user applications remains challenging. DL algorithms are often computationally expensive, power-hungry, and require large memory to process complex and iterative operations of millions of parameters. Hence, training and inference of DL models are typically performed on high-performance computing (HPC) clusters in the cloud. Data transmission to the cloud results in high latency, round-trip delay, security and privacy concerns, and the inability of real-time decisions. Thus, processing on edge devices can significantly reduce cloud transmission cost. Edge devices are end devices closest to the user, such as mobile phones, cyber-physical systems (CPSs), wearables, the Internet of Things (IoT), embedded and autonomous systems, and intelligent sensors. These devices have limited memory, computing resources, and power-handling capability. Therefore, optimization techniques at both the hardware and software levels have been developed to handle the DL deployment efficiently on the edge. Understanding the existing research, challenges, and opportunities is fundamental to leveraging the next generation of edge devices with artificial intelligence (AI) capability. Mainly, four research directions have been pursued for efficient DL inference on edge devices: 1) novel DL architecture and algorithm design; 2) optimization of existing DL methods; 3) development of algorithm-hardware codesign; and 4) efficient accelerator design for DL deployment. This article focuses on surveying each of the four research directions, providing a comprehensive review of the state-of-the-art tools and techniques for efficient edge inference.
Author	Shuvo, Md. Maruf Hossain Morshed, Bashir I. Cheng, Jianlin Islam, Syed Kamrul
Author_xml	– sequence: 1 givenname: Md. Maruf Hossain orcidid: 0000-0002-3498-4947 surname: Shuvo fullname: Shuvo, Md. Maruf Hossain organization: Department of Electrical Engineering and Computer Science, Analog/Mixed Signal VLSI and Devices Laboratory (AVDL), University of Missouri, Columbia, MO, USA – sequence: 2 givenname: Syed Kamrul orcidid: 0000-0002-0501-0027 surname: Islam fullname: Islam, Syed Kamrul organization: Department of Electrical Engineering and Computer Science, Analog/Mixed Signal VLSI and Devices Laboratory (AVDL), University of Missouri, Columbia, MO, USA – sequence: 3 givenname: Jianlin surname: Cheng fullname: Cheng, Jianlin organization: Department of Electrical Engineering and Computer Science, Bioinformatics and Machine Learning Laboratory (BML), University of Missouri, Columbia, MO, USA – sequence: 4 givenname: Bashir I. orcidid: 0000-0002-2178-433X surname: Morshed fullname: Morshed, Bashir I. organization: Department of Computer Science, Cyber Physical Systems (CPS) Laboratory, Texas Tech University, Lubbock, TX, USA
BookMark	eNp9kMtOwzAQRS0EEm3hB2ATiXXK2HlN2FWhQFGlogrWwXHGlaviFDst4u9JH2LBgtUs5p55nD47tY0lxq44DDmH_Pb5ZT4rhgKEGEZCpDHyE9bjSYKhEEl6ynoAHMNc8Pyc9b1fAkCUpFGPvY-1NsqQbYORUrQiJ1vT2KDRwT3ROpiSdNbYRTCxmhxZRUHXnZNvNk5RWDTWt04aS3UwrhfUQVujyN8Foy60NfR1wc60XHm6PNYBe3sYvxZP4XT2OClG01DFkLQhRw51JutY6Ayx6u4DjVyoSmshBaYVZjKmREecK53VkkQudR1VCJBBpTAasJvD3LVrPjfk23LZnWi7laXI0gghR8y7lDiklGu8d6TLtTMf0n2XHMqdyXJvstyZLI8mOwj_QMq0e02711f_o9cH1BDR7648xwQAox8xu4MF
CODEN	IEEPAD
CitedBy_id	crossref_primary_10_1109_TII_2024_3396559 crossref_primary_10_1109_TCAD_2024_3355190 crossref_primary_10_1109_JBHI_2022_3233486 crossref_primary_10_1109_ACCESS_2025_3544515 crossref_primary_10_1016_j_measurement_2025_117618 crossref_primary_10_3390_iot6020029 crossref_primary_10_1109_ACCESS_2025_3599356 crossref_primary_10_1049_cdt2_6214436 crossref_primary_10_1016_j_physa_2024_129909 crossref_primary_10_1016_j_engappai_2025_112241 crossref_primary_10_1515_auto_2024_0155 crossref_primary_10_1109_ACCESS_2024_3485550 crossref_primary_10_1109_ACCESS_2025_3565296 crossref_primary_10_1007_s10489_024_05615_7 crossref_primary_10_1109_JIOT_2024_3456569 crossref_primary_10_3390_fi17100433 crossref_primary_10_3390_agronomy15051128 crossref_primary_10_1093_jas_skae098 crossref_primary_10_1109_ACCESS_2025_3562612 crossref_primary_10_1007_s11554_025_01670_6 crossref_primary_10_1109_ACCESS_2024_3452184 crossref_primary_10_1016_j_rser_2025_116231 crossref_primary_10_1016_j_nanoen_2024_110186 crossref_primary_10_3390_computers14090374 crossref_primary_10_1007_s00607_024_01377_9 crossref_primary_10_1016_j_knosys_2025_113882 crossref_primary_10_1007_s11053_025_10475_0 crossref_primary_10_1109_TII_2024_3507954 crossref_primary_10_3389_fcomp_2025_1538277 crossref_primary_10_1016_j_jfca_2025_108161 crossref_primary_10_1007_s11760_025_04219_z crossref_primary_10_3390_agriculture15121238 crossref_primary_10_1109_TPDS_2024_3469545 crossref_primary_10_1109_TMC_2024_3465434 crossref_primary_10_1109_ACCESS_2024_3358827 crossref_primary_10_1016_j_atech_2025_101330 crossref_primary_10_1109_TVT_2024_3395292 crossref_primary_10_1016_j_jhazmat_2024_134723 crossref_primary_10_1109_JSEN_2025_3583820 crossref_primary_10_1016_j_comnet_2025_111592 crossref_primary_10_3390_agriculture15151643 crossref_primary_10_3390_bioengineering10111324 crossref_primary_10_1007_s10462_024_10774_7 crossref_primary_10_1007_s10586_025_05226_y crossref_primary_10_1109_ACCESS_2025_3608445 crossref_primary_10_1145_3768630 crossref_primary_10_3389_fcomp_2025_1563942 crossref_primary_10_3389_fnins_2025_1656892 crossref_primary_10_1016_j_jnca_2025_104231 crossref_primary_10_1109_TCAD_2024_3448379 crossref_primary_10_1016_j_suscom_2024_101004 crossref_primary_10_1109_ACCESS_2023_3276438 crossref_primary_10_1016_j_dsp_2024_104851 crossref_primary_10_3390_s24185965 crossref_primary_10_1016_j_compbiolchem_2025_108620 crossref_primary_10_3390_agriculture15151653 crossref_primary_10_1007_s10462_024_11033_5 crossref_primary_10_1016_j_cmpbup_2025_100205 crossref_primary_10_1038_s41598_025_15351_8 crossref_primary_10_1007_s00607_025_01427_w crossref_primary_10_1038_s41598_025_16043_z crossref_primary_10_1016_j_lfs_2025_123931 crossref_primary_10_1016_j_eswa_2025_128427 crossref_primary_10_3390_info15030161 crossref_primary_10_1109_JSEN_2024_3361980 crossref_primary_10_1109_TSC_2025_3586152 crossref_primary_10_3390_atmos15060689 crossref_primary_10_1109_JPROC_2024_3437365 crossref_primary_10_32604_cmc_2024_057353 crossref_primary_10_1016_j_comnet_2025_111654 crossref_primary_10_1016_j_fraope_2025_100256 crossref_primary_10_1016_j_neucom_2025_130097 crossref_primary_10_1109_JSEN_2024_3502539 crossref_primary_10_1109_TSC_2024_3359148 crossref_primary_10_1109_JIOT_2025_3542534 crossref_primary_10_2298_CSIS240503020S crossref_primary_10_1371_journal_pone_0323689 crossref_primary_10_1016_j_patcog_2024_110749 crossref_primary_10_1016_j_sysarc_2025_103409 crossref_primary_10_1109_TII_2024_3353851 crossref_primary_10_3390_electronics12132908 crossref_primary_10_32628_IJSRST2513120 crossref_primary_10_1109_TCAD_2024_3443694 crossref_primary_10_3390_s25123784 crossref_primary_10_1109_JIOT_2025_3546577 crossref_primary_10_1109_JSTARS_2024_3424954 crossref_primary_10_1587_transele_2024LHP0004 crossref_primary_10_3390_s24165262 crossref_primary_10_3390_su17125255 crossref_primary_10_1145_3724420 crossref_primary_10_3390_plants14101481 crossref_primary_10_21833_ijaas_2025_09_003 crossref_primary_10_1016_j_sysarc_2025_103536 crossref_primary_10_1016_j_aei_2025_103516 crossref_primary_10_1007_s00500_023_09109_5 crossref_primary_10_1016_j_iot_2025_101488 crossref_primary_10_1109_TCSI_2024_3446582 crossref_primary_10_1016_j_aej_2025_08_030 crossref_primary_10_1109_TIM_2024_3364264 crossref_primary_10_3390_s24175586 crossref_primary_10_1007_s11831_023_10032_z crossref_primary_10_1109_JIOT_2024_3483232 crossref_primary_10_1016_j_cose_2024_103733 crossref_primary_10_3390_s24092807 crossref_primary_10_1016_j_neucom_2025_131163 crossref_primary_10_3390_s23146485 crossref_primary_10_3390_math13111878 crossref_primary_10_1109_TCSI_2024_3369230 crossref_primary_10_1007_s10846_024_02141_z crossref_primary_10_1111_exsy_13760 crossref_primary_10_3390_s25010083 crossref_primary_10_3390_sym17060891 crossref_primary_10_1097_ICU_0000000000000985 crossref_primary_10_1109_TIM_2025_3581658 crossref_primary_10_3390_s24155069 crossref_primary_10_1109_TCE_2023_3339468 crossref_primary_10_3390_electronics14102038
Cites_doi	10.1109/PerCom45495.2020.9127387 10.1109/CVPR.2016.308 10.1007/978-3-642-27645-3 10.1109/JIOT.2020.2967734 10.3390/a13050125 10.1109/TPAMI.2013.50 10.1109/JPROC.2021.3098483 10.1109/HPCA.2017.29 10.1145/3038912.3052577 10.1145/1143844.1143865 10.1109/TVLSI.2019.2939429 10.1109/IEMTE.1998.723053 10.1109/ISCA.2018.00062 10.1109/IJCNN.2019.8852463 10.1145/3089801.3089803 10.1007/978-3-031-14748-7_2 10.1109/ICCV.2017.298 10.1145/3241539.3241557 10.1109/ICCE.2016.7430525 10.1145/3152127 10.1006/ijhc.2001.0499 10.1109/VLSIC.2018.8502438 10.1109/ISCA.2018.00040 10.1109/CODESISSS.2015.7331375 10.1109/TPAMI.2021.3055564 10.1109/JPROC.2019.2918951 10.1145/2810103.2813687 10.14778/3137628.3137664 10.1145/3309551 10.1145/3140659.3080215 10.1109/VLSIC.2018.8502276 10.1109/TPAMI.2018.2858232 10.1145/3377713.3377721 10.3390/electronics10091025 10.1007/978-3-030-37334-4_2 10.1145/3081333.3081336 10.1038/s41586-019-1424-8 10.18653/v1/2022.emnlp-main.741 10.1145/3007787.3001138 10.1007/978-3-030-05318-5_3 10.1145/2872887.2750389 10.1109/ICCV.2019.00140 10.1109/ISSCC.2014.6757323 10.1109/IPSN.2016.7460664 10.1109/ACCESS.2021.3131396 10.1109/JPROC.2021.3119950 10.1145/2554688.2554785 10.1109/MeMeA54994.2022.9856420 10.1109/PACT.2019.00009 10.1145/3093337.3037698 10.1109/APCCAS.2018.8605639 10.1109/IEDM.2017.8268341 10.1109/VLSID.2019.00055 10.3390/fi12070113 10.1109/MICRO.2014.58 10.1145/3218603.3218643 10.1109/MICRO.2018.00062 10.1073/pnas.1604850113 10.1109/ISCA.2018.00060 10.1109/ISCA.2018.00012 10.1145/3287624.3287627 10.1109/PDP2018.2018.00023 10.1109/JSSC.2017.2778281 10.1109/DAC.2018.8465915 10.1007/978-3-319-46493-0_32 10.1109/TPDS.2020.3030548 10.1609/aaai.v34i04.5963 10.1109/ICDCS.2017.226 10.1145/3089801.3089804 10.1145/2627369.2627613 10.1016/j.pmcj.2017.07.014 10.23919/VLSIC.2019.8778056 10.1109/TIFS.2015.2400395 10.1109/ISNE.2017.7968748 10.1109/CVPR.2016.90 10.1109/UCC48980.2020.00066 10.1145/3089801.3089806 10.1145/3079856.3080246 10.1145/2906388.2906396 10.1109/CVPR.2019.01147 10.1145/3161174 10.1109/JETCAS.2019.2910232 10.1109/ASPDAC.2017.7858419 10.1109/MICRO.2016.7783723 10.3233/JIFS-169699 10.1109/ICCCN.2018.8487352 10.23919/VLSIC.2019.8778006 10.1109/I2MTC50364.2021.9459794 10.1109/MWC.006.2100699 10.1109/EEEI.2014.7005803 10.1109/ISVLSI.2016.111 10.1145/3240765.3240801 10.1109/MSP.2017.2765695 10.1145/3352460.3358269 10.23919/VLSIC.2019.8778193 10.1145/3089801.3089805 10.1109/MM.2018.053631145 10.1109/ISCA.2018.00069 10.1016/j.neucom.2017.09.046 10.1109/FCCM.2017.64 10.1109/CVPRW.2018.00215 10.1109/ICDCS.2017.94 10.1016/j.sysarc.2020.101887 10.1609/aimag.v35i4.2513 10.1609/aaai.v32i1.11601 10.1109/ICASSP.2017.7953288 10.1145/3132211.3134456 10.1109/TC.2019.2914438 10.1145/3154448.3154452 10.1109/ISCA.2018.00068 10.1145/3174243.3174253 10.1109/BTAS46853.2019.9185981 10.3390/electronics5040088 10.3390/s21092984 10.1145/3081333.3081359 10.1145/3079856.3080244 10.1109/ICPR.2018.8546129 10.1109/5.726791 10.3390/s18113726 10.1109/MDAT.2017.2741463 10.1145/2994551.2994564 10.1145/3131895 10.1145/3242897 10.2200/s00832ed1v01y201802aim037 10.1109/DAC.2018.8465793 10.1145/3494834.3500240 10.1109/ISCA.2018.00061 10.18653/v1/2020.acl-main.686 10.1109/ISQED54688.2022.9806197 10.1145/2996864 10.1145/3229556.3229562 10.1109/ACCESS.2018.2870052 10.1145/3297858.3304028 10.1109/FCCM.2017.47 10.1145/3373376.3378534 10.1145/3287624.3287715 10.48550/arXiv.1503.02531 10.1109/CVPR.2018.00171 10.1007/978-3-030-27562-4_6 10.1145/3007787.3001163 10.1109/HPCA.2019.00027 10.1109/EMC2-NIPS53020.2019.00013 10.1145/2820975.2820980 10.1145/2935643.2935650 10.1007/978-3-030-01261-8_25 10.1109/SmartWorld-UIC-ATC-SCALCOM-IOP-SCI.2019.00194 10.1145/3289602.3293902 10.1109/ICCD.2017.49 10.1109/TNET.2020.3042320 10.1109/JPROC.2019.2921977 10.1145/3140659.3080254 10.1109/TC.2019.2924215 10.1109/ICTAI.2019.00197 10.1145/2647868.2654889 10.1145/3007787.3001140 10.1145/3352460.3358295 10.1609/aaai.v31i1.11231 10.23919/VLSIC.2017.8008533 10.3390/electronics8030292 10.18653/v1/2020.emnlp-demos.6 10.1145/3299710.3211336 10.1109/CVPR.2018.00474 10.1109/ICPR.2018.8545462 10.1145/3267809.3267828 10.1145/3093336.3037746 10.1109/AIPR50011.2020.9425332 10.1109/ACCESS.2018.2874823 10.1109/CASES.2015.7324548 10.1109/CVPR.2016.91 10.1609/aaai.v33i01.33013779 10.1109/TNNLS.2018.2852335 10.1609/aaai.v32i1.11876 10.1109/LCA.2016.2597140 10.1145/3472291 10.1111/1754-9485.13261 10.1038/s41586-018-0180-5 10.1109/ISSCC.2019.8662302 10.1109/ICASSP.2015.7178146 10.1007/s11633-016-1006-2 10.1145/3289602.3293905 10.1007/978-0-387-84858-7_14 10.1109/WACV.2018.00083 10.1109/ICCCBDA51879.2021.9442600 10.1109/CVPR.2019.00293 10.21437/Interspeech.2021-1816 10.1109/ISLPED.2017.8009173 10.1145/3352460.3358291 10.1007/978-1-4302-4000-6_9 10.1609/aaai.v34i04.5954 10.1109/ICPR.2016.7900006 10.23919/DATE.2017.7927211 10.1007/s00521-018-3761-1 10.1609/aaai.v35i2.16179 10.1109/FCCM.2019.00033 10.1109/VLSIC.2018.8502404 10.1109/ISCAS51556.2021.9401083 10.1109/JPROC.2017.2761740 10.1145/3109761.3109804 10.1109/TIM.2020.3018831 10.1086/719650 10.1007/s10994-019-05855-6 10.7551/mitpress/12832.003.0015 10.23919/DATE48585.2020.9116529 10.1109/SEC.2016.38 10.1109/LCA.2020.2979965 10.1145/3318216.3363316 10.1109/CVPR.2018.00454 10.1109/CVPR.2017.243 10.1007/978-3-030-01249-6_18 10.1109/ISSCC.2018.8310262 10.4108/eai.30-11-2016.2267463 10.1109/JSSC.2016.2616357 10.1109/TCAD.2018.2858384 10.1613/jair.301 10.3390/en15207495 10.1109/ACCESS.2018.2890150 10.1109/CVPR.2016.435 10.1109/CVPR42600.2020.01464 10.1145/3174243.3174261 10.1007/978-3-030-01246-5_2 10.1109/VLSIC.2016.7573525 10.1145/3296957.3173176 10.1109/PERCOMW.2016.7457169 10.1109/EMC2.2018.00012 10.1145/3020078.3021745 10.1109/CVPRW.2011.5981829 10.5244/C.28.88 10.1109/LES.2018.2815954 10.1109/MPRV.2017.2940968 10.4324/9781410605337-29 10.1109/LCOMM.2019.2921755 10.18653/v1/2020.findings-emnlp.372 10.1145/3020078.3021740 10.1016/j.micpro.2020.102991 10.1109/CVPR.2018.00745 10.23919/VLSIC.2017.8008534 10.1145/2750858.2804262 10.1109/FPL.2018.00035 10.1109/JETCAS.2018.2835809 10.1109/TNNLS.2018.2844093 10.1145/3287624.3287714 10.1109/MICRO.2018.00024 10.1109/CVPR.2017.574 10.1145/2964284.2973801 10.1109/CVPR.2018.00716 10.1109/JPROC.2020.2976475 10.1109/PADSW.2018.8645013 10.1109/CVPR.2017.195 10.1145/3289602.3293910 10.1109/MICRO.2016.7783720 10.1109/INFOCOM.2018.8486403 10.3390/electronics8111321 10.2200/s00960ed2v01y201910aim043 10.1109/MICRO.2018.00011 10.1007/978-3-030-01267-0_44 10.1007/978-3-319-97909-0_46 10.1109/EMBC.2017.8036767 10.1109/CVPR42600.2020.00396 10.1145/3418297 10.1109/ICASSP.2016.7472657 10.1109/IPDPS53621.2022.00110 10.1145/1150402.1150464 10.1109/TII.2018.2842821 10.48550/arXiv.1810.04805 10.1109/FPL.2018.00075 10.1145/3234150 10.1016/j.patcog.2015.03.009 10.1145/3212725.3212728 10.1109/ICPADS.2017.00069 10.1109/CHASE.2017.115 10.1109/ISSCC.2017.7870353 10.1109/INFOCOM48880.2022.9796896 10.1145/3081333.3081360 10.1109/ICCVW.2019.00447 10.1109/CVPR.2017.643 10.1137/1.9781611970364 10.1109/TVLSI.2018.2797600 10.1145/3240765.3243473 10.1109/CVPR.2018.00291 10.1109/ICACSIS.2017.8355051 10.1007/978-3-540-75171-7_2 10.1145/3289602.3293898 10.1109/TCAD.2017.2705069 10.1109/TMC.2020.2999956
ContentType	Journal Article
Copyright	Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023
Copyright_xml	– notice: Copyright The Institute of Electrical and Electronics Engineers, Inc. (IEEE) 2023
DBID	97E ESBDL RIA RIE AAYXX CITATION 7SP 8FD L7M
DOI	10.1109/JPROC.2022.3226481
DatabaseName	IEEE All-Society Periodicals Package (ASPP) 2005–Present IEEE Xplore Open Access Journals IEEE All-Society Periodicals Package (ASPP) 1998–Present IEEE/IET Electronic Library CrossRef Electronics & Communications Abstracts Technology Research Database Advanced Technologies Database with Aerospace
DatabaseTitle	CrossRef Technology Research Database Advanced Technologies Database with Aerospace Electronics & Communications Abstracts
DatabaseTitleList	Technology Research Database
Database_xml	– sequence: 1 dbid: RIE name: IEEE/IET Electronic Library url: https://ieeexplore.ieee.org/ sourceTypes: Publisher
DeliveryMethod	fulltext_linktorsrc
Discipline	Engineering
EISSN	1558-2256
EndPage	50
ExternalDocumentID	10_1109_JPROC_2022_3226481 9985008
Genre	orig-research
GrantInformation_xml	– fundername: College of Engineering, University of Missouri, Columbia, MO, USA
GroupedDBID	-DZ -~X .DC 0R~ 123 29P 4.4 6IK 85S 97E AAJGR AAWTH ABAZT ABJNI ABQJQ ABVLG ACBEA ACGFS AENEX AFOGA AGQYO AGSQL AHBIQ ALMA_UNASSIGNED_HOLDINGS AZLTO BEFXN BFFAM BGNUA BKEBE BPEOZ CS3 DU5 EBS ESBDL HZ~ IFIPE IPLJI JAVBF LAI M43 O9- OCL RIA RIE RIU RNS TAE TN5 TWZ UHB UKR UQL YNT ZCA ~02 1OL 3EH 9M8 AAYXX ABFSI AETEA AETIX AGNAY AIBXA ALLEH CITATION EJD FA8 H~9 IAAWW IBMZZ ICLAB IDIHD IFJZH MVM VOH WHG XOL ZXP ZY4 7SP 8FD L7M
ID	FETCH-LOGICAL-c405t-1810d7ad42f788b0000f812cbff2a286b87a4e5f311cf7dae29afd3b80070bc83
IEDL.DBID	RIE
ISICitedReferencesCount	174
ISICitedReferencesURI	http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000899996300001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN	0018-9219
IngestDate	Sun Nov 09 08:26:08 EST 2025 Sat Nov 29 06:01:44 EST 2025 Tue Nov 18 22:20:17 EST 2025 Wed Aug 27 02:29:16 EDT 2025
IsDoiOpenAccess	true
IsOpenAccess	true
IsPeerReviewed	true
IsScholarly	true
Issue	1
Language	English
License	https://creativecommons.org/licenses/by/4.0/legalcode
LinkModel	DirectLink
MergedId	FETCHMERGED-LOGICAL-c405t-1810d7ad42f788b0000f812cbff2a286b87a4e5f311cf7dae29afd3b80070bc83
Notes	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ORCID	0000-0002-3498-4947 0000-0002-0501-0027 0000-0002-2178-433X
OpenAccessLink	https://ieeexplore.ieee.org/document/9985008
PQID	2763809889
PQPubID	85453
PageCount	50
ParticipantIDs	crossref_primary_10_1109_JPROC_2022_3226481 proquest_journals_2763809889 ieee_primary_9985008 crossref_citationtrail_10_1109_JPROC_2022_3226481
PublicationCentury	2000
PublicationDate	2023-01-01
PublicationDateYYYYMMDD	2023-01-01
PublicationDate_xml	– month: 01 year: 2023 text: 2023-01-01 day: 01
PublicationDecade	2020
PublicationPlace	New York
PublicationPlace_xml	– name: New York
PublicationTitle	Proceedings of the IEEE
PublicationTitleAbbrev	JPROC
PublicationYear	2023
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	ref57 Maaz (ref71) 2022 ref207 ref328 ref56 ref208 ref329 ref326 ref58 ref327 ref53 ref203 ref324 ref52 ref325 ref55 ref201 ref322 ref54 ref323 Sak (ref89) 2014 ref209 ref210 ref331 ref332 Shacham (ref330) 2022 ref218 ref339 ref219 ref216 ref337 ref217 ref338 ref42 ref214 ref335 ref215 ref336 ref212 ref333 ref43 Mao (ref365) Pan (ref298) 2020 (ref266) 2022 ref49 Courbariaux (ref134) 2016 ref100 ref342 ref101 ref222 ref343 ref340 ref220 ref341 Lin (ref115); 48 ref34 ref37 ref304 ref36 ref423 ref424 ref421 ref422 Han (ref157) 2015 ref39 ref38 ref308 ref309 ref310 Ogden (ref370) ref317 ref23 ref318 ref26 ref315 ref25 Wang (ref193) 2018 ref316 ref20 ref313 ref314 ref22 ref311 ref21 (ref253) 2022 ref28 ref27 Zagoruyko (ref205) 2016 ref319 ref320 ref200 ref321 Sergeev (ref297) 2018 Esser (ref136) ref128 Hubara (ref142) 2016 ref249 ref129 ref97 ref126 ref247 ref96 ref248 ref369 ref124 ref245 ref125 ref246 ref367 Howard (ref70) 2017 Kim (ref175) 2015 ref375 Buchlovsky (ref301) 2019 ref376 ref95 ref131 ref252 ref373 ref374 ref250 ref371 ref130 ref251 ref372 McMahan (ref359) 2017 Konečný (ref363) 2016 ref139 (ref62) 2022 ref258 ref379 ref85 ref138 ref259 ref256 ref377 ref87 ref378 Zhang (ref334) 2017 ref82 ref144 ref386 ref81 ref387 ref384 ref83 ref385 ref140 ref261 ref262 ref383 ref380 Hojabr (ref180) ref381 ref79 ref108 Chen (ref149) ref78 ref109 Shazeer (ref300) ref348 ref107 ref349 ref75 ref104 ref346 ref74 ref226 ref347 ref223 ref344 ref76 ref103 ref224 (ref272) 2022 Molchanov (ref105) 2016 Lai (ref159) 2018 Hubara (ref141); 4107 ref111 ref232 ref353 ref233 ref354 ref230 ref351 Harlap (ref368) 2018 ref72 ref110 ref231 ref352 Gupta (ref116); 37 ref350 ref68 ref119 ref67 ref117 ref238 Kumar (ref31); 70 ref118 ref239 (ref268) 2022 ref64 ref236 ref357 Wendelken (ref3) 2022 ref237 ref358 ref66 ref234 ref355 ref65 ref235 ref356 Brown (ref46) 2020 Huang (ref296) Zhang (ref92) ref60 ref122 ref243 ref364 ref123 ref244 Narang (ref228) 2017 ref241 ref362 ref121 ref242 ref360 ref240 ref361 Baluja (ref165) 2018 Li (ref41) 2018; abs/1810.06339 ref168 ref289 Kalchbrenner (ref93) 2015 ref169 ref177 ref178 ref173 ref174 ref171 ref179 He (ref94) ref181 ref188 Zoph (ref61) 2016 ref186 Dai (ref302) ref187 (ref269) 2022 ref182 ref183 Han (ref47) 2015 Neil (ref88) ref388 Zoph (ref59) 2017 ref147 ref389 Krizhevsky (ref50); 25 Zhou (ref143) 2016 ref390 Guihot (ref279) 2012 Gupta (ref29) ref276 ref397 Gong (ref163) 2014 ref277 ref398 ref395 ref275 ref396 ref151 ref393 ref273 ref394 ref391 ref150 ref392 Wen (ref98); 29 ref278 ref399 Li (ref156) 2022 Lin (ref133) 2015 ref280 ref287 ref167 ref288 ref164 ref285 ref286 ref162 ref283 ref284 ref160 ref281 ref161 Yao (ref295) 2022 Liu (ref274) 2020 Soudry (ref135) Vanhoucke (ref120) Vaswani (ref45) Gale (ref225) 2019 Sarwar (ref166) ref2 ref1 Moritz (ref303) ref191 ref192 ref190 ref199 ref197 ref195 ref196 Courbariaux (ref132) Denton (ref202) Chen (ref221) 2017 Lebedev (ref176) 2014 Wen (ref410) 2020 Zhang (ref170) Xu (ref172) 2017 Zhou (ref152) 2017 Rasley (ref290) Wang (ref154) (ref260) 2022 Bohdal (ref194) 2020 Rallapalli (ref282) 2016 Lin (ref264); 33 ref8 ref7 ref9 ref4 Baines (ref299) 2021 ref6 ref5 Guan (ref263) 2021; 64 Agostinelli (ref44) 2014 Zhu (ref229) 2017 (ref265) 2022 Wang (ref227) 2020 Kuchaiev (ref90) 2017 Mao (ref99) 2017 Xu (ref153) 2018 Cai (ref418) 2018 Zhu (ref146) 2016 Li (ref145) 2016 Crowley (ref204) Romero (ref185) 2014 Srivastava (ref48) 2014; 15 (ref267) 2022 Lu (ref293) 2022 Refaeilzadeh (ref33) 2020; 5 Micikevicius (ref112) 2017 (ref271) 2022 Bie (ref155) 2019 Meng (ref80) 2019 van der Westhuizen (ref86) 2018 Liu (ref158) Iandola (ref73) 2016 You (ref106) 2019 Ruder (ref35) 2016 Xiong (ref77) 2019 Wu (ref91) 2016 (ref63) 2022 Jelčicová (ref102) 2022 Yang (ref345) (ref306) 2022 Simonyan (ref51) 2014 (ref270) 2021 Chaulwar (ref189) 2022 Courbariaux (ref137) 2016 Jozefowicz (ref84) Tann (ref211) Bolukbasi (ref213) 2017 Shafiee (ref198) 2017 Xu (ref24) 2017 Gysel (ref114) Zhu (ref40) 2005 Wang (ref366) Liu (ref69) 2018 (ref305) 2022 Ryu (ref127) Abadi (ref255) Li (ref294) 2021 ref13 ref405 ref12 ref406 ref15 ref403 ref14 ref404 ref401 ref402 ref11 ref10 ref400 (ref312) 2022 ref17 ref409 ref16 ref19 Ren (ref291) ref407 ref18 ref408 Wang (ref113) 2018 ref416 ref417 ref414 ref415 ref412 ref413 ref411 Covell (ref148) 2019 ref419 (ref307) 2022 (ref254) 2022 ref420 Dennis (ref30) 2020 Paszke (ref257); 32 Murphy (ref32) 2012 Ge (ref382) 2022 Sau (ref206) 2016 Huang (ref184) 2017 Tang (ref292)
References_xml	– ident: ref384 doi: 10.1109/PerCom45495.2020.9127387 – ident: ref54 doi: 10.1109/CVPR.2016.308 – year: 2018 ident: ref153 article-title: Alternating multi-bit quantization for recurrent neural networks publication-title: arXiv:1802.00150 – ident: ref43 doi: 10.1007/978-3-642-27645-3 – ident: ref367 doi: 10.1109/JIOT.2020.2967734 – year: 2022 ident: ref382 article-title: Lossless acceleration for Seq2seq generation with aggressive decoding publication-title: arXiv:2205.10350 – ident: ref16 doi: 10.3390/a13050125 – volume-title: Stratus High-Level Synthesis year: 2022 ident: ref271 – ident: ref412 doi: 10.1109/TPAMI.2013.50 – ident: ref15 doi: 10.1109/JPROC.2021.3098483 – ident: ref311 doi: 10.1109/HPCA.2017.29 – year: 2022 ident: ref189 article-title: Extreme compression of sentence-transformer ranker models: Faster inference, longer battery life, and less storage on edge devices publication-title: arXiv:2207.12852 – year: 2017 ident: ref229 article-title: To prune, or not to prune: Exploring the efficacy of pruning for model compression publication-title: arXiv:1710.01878 – ident: ref392 doi: 10.1145/3038912.3052577 – ident: ref37 doi: 10.1145/1143844.1143865 – year: 2018 ident: ref165 article-title: No multiplication? No floating point? No problem! Training networks for efficient inference publication-title: arXiv:1809.09244 – ident: ref407 doi: 10.1109/TVLSI.2019.2939429 – volume-title: Tensilica DNA Processor Family for on-Device AI year: 2022 ident: ref305 – ident: ref331 doi: 10.1109/IEMTE.1998.723053 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref296 article-title: GPipe: Efficient training of giant neural networks using pipeline parallelism – year: 2018 ident: ref193 article-title: Dataset distillation publication-title: arXiv:1811.10959 – ident: ref251 doi: 10.1109/ISCA.2018.00062 – ident: ref107 doi: 10.1109/IJCNN.2019.8852463 – ident: ref399 doi: 10.1145/3089801.3089803 – ident: ref351 doi: 10.1007/978-3-031-14748-7_2 – year: 2016 ident: ref73 article-title: SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5 MB model size publication-title: arXiv:1602.07360 – year: 2014 ident: ref176 article-title: Speeding-up convolutional neural networks using fine-tuned CP-decomposition publication-title: arXiv:1412.6553 – volume: 70 start-page: 1935 volume-title: Proc. 34th Int. Conf. Mach. Learn. ident: ref31 article-title: Resource-efficient machine learning in 2 KB RAM for the Internet of Things – ident: ref195 doi: 10.1109/ICCV.2017.298 – ident: ref23 doi: 10.1145/3241539.3241557 – year: 2019 ident: ref225 article-title: The state of sparsity in deep neural networks publication-title: arXiv:1902.09574 – year: 2016 ident: ref105 article-title: Pruning convolutional neural networks for resource efficient inference publication-title: arXiv:1611.06440 – ident: ref207 doi: 10.1109/ICCE.2016.7430525 – ident: ref394 doi: 10.1145/3152127 – ident: ref356 doi: 10.1006/ijhc.2001.0499 – year: 2017 ident: ref112 article-title: Mixed precision training publication-title: arXiv:1710.03740 – ident: ref327 doi: 10.1109/VLSIC.2018.8502438 – ident: ref338 doi: 10.1109/ISCA.2018.00040 – year: 2019 ident: ref148 article-title: Table-based neural units: Fully quantizing networks for multiply-free inference publication-title: arXiv:1906.04798 – ident: ref210 doi: 10.1109/CODESISSS.2015.7331375 – ident: ref178 doi: 10.1109/TPAMI.2021.3055564 – ident: ref18 doi: 10.1109/JPROC.2019.2918951 – ident: ref362 doi: 10.1145/2810103.2813687 – start-page: 1 volume-title: Proc. Int. Conf. Learn. Represent. ident: ref114 article-title: Hardware-oriented approximation of convolutional neural networks – ident: ref209 doi: 10.14778/3137628.3137664 – ident: ref7 doi: 10.1145/3309551 – ident: ref101 doi: 10.1145/3140659.3080215 – ident: ref318 doi: 10.1109/VLSIC.2018.8502276 – year: 2016 ident: ref91 article-title: Google’s neural machine translation system: Bridging the gap between human and machine translation publication-title: arXiv:1609.08144 – ident: ref405 doi: 10.1109/TPAMI.2018.2858232 – ident: ref138 doi: 10.1145/3377713.3377721 – ident: ref17 doi: 10.3390/electronics10091025 – ident: ref258 doi: 10.1007/978-3-030-37334-4_2 – ident: ref395 doi: 10.1145/3081333.3081336 – ident: ref324 doi: 10.1038/s41586-019-1424-8 – ident: ref381 doi: 10.18653/v1/2022.emnlp-main.741 – ident: ref245 doi: 10.1145/3007787.3001138 – volume-title: Pixel Visual Core: Image Processing and Machine Learning on Pixel 2 year: 2022 ident: ref330 – volume: 33 start-page: 11711 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref264 article-title: MCUNet: Tiny deep learning on IoT devices – year: 2020 ident: ref410 article-title: Time series data augmentation for deep learning: A survey publication-title: arXiv:2002.12478 – volume: abs/1810.06339 year: 2018 ident: ref41 article-title: Deep reinforcement learning publication-title: CoRR – ident: ref60 doi: 10.1007/978-3-030-05318-5_3 – start-page: 1331 volume-title: Proc. Int. Conf. Mach. Learn. ident: ref29 article-title: ProtoNN: Compressed and accurate KNN for resource-scarce devices – ident: ref286 doi: 10.1145/2872887.2750389 – ident: ref64 doi: 10.1109/ICCV.2019.00140 – ident: ref118 doi: 10.1109/ISSCC.2014.6757323 – ident: ref284 doi: 10.1109/IPSN.2016.7460664 – year: 2017 ident: ref228 article-title: Exploring sparsity in recurrent neural networks publication-title: arXiv:1704.05119 – ident: ref377 doi: 10.1109/ACCESS.2021.3131396 – ident: ref20 doi: 10.1109/JPROC.2021.3119950 – ident: ref247 doi: 10.1145/2554688.2554785 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref94 article-title: Wider and deeper, cheaper and faster: Tensorized LSTMs for sequence learning – ident: ref27 doi: 10.1109/MeMeA54994.2022.9856420 – start-page: 1 volume-title: Proc. USENIX Workshop Hot Topics Edge Comput. (HotEdge) ident: ref365 article-title: A privacy-preserving deep learning approach for face recognition with edge computing – year: 2014 ident: ref185 article-title: FitNets: Hints for thin deep nets publication-title: arXiv:1412.6550 – ident: ref230 doi: 10.1109/PACT.2019.00009 – ident: ref25 doi: 10.1145/3093337.3037698 – year: 2016 ident: ref206 article-title: Deep model compression: Distilling knowledge from noisy teachers publication-title: arXiv:1610.09650 – ident: ref328 doi: 10.1109/APCCAS.2018.8605639 – ident: ref340 doi: 10.1109/IEDM.2017.8268341 – ident: ref82 doi: 10.1109/VLSID.2019.00055 – start-page: 1 volume-title: Proc. 56th ACM/IEEE Design Automat. Conf. (DAC) ident: ref180 article-title: SkippyNN: An embedded stochastic-computing accelerator for convolutional neural networks – year: 2015 ident: ref47 article-title: Learning both weights and connections for efficient neural networks publication-title: arXiv:1506.02626 – start-page: 2342 volume-title: Proc. Int. Conf. Mach. Learn. ident: ref84 article-title: An empirical exploration of recurrent network architectures – year: 2016 ident: ref143 article-title: DoReFa-Net: Training low bitwidth convolutional neural networks with low bitwidth gradients publication-title: arXiv:1606.06160 – ident: ref8 doi: 10.3390/fi12070113 – volume-title: Intel® Edison Development Platform year: 2022 ident: ref254 – ident: ref326 doi: 10.1109/MICRO.2014.58 – ident: ref173 doi: 10.1145/3218603.3218643 – volume: 37 start-page: 1737 volume-title: Proc. 32nd Int. Conf. Mach. Learn. (ICML) ident: ref116 article-title: Deep learning with limited numerical precision – ident: ref339 doi: 10.1109/MICRO.2018.00062 – ident: ref147 doi: 10.1073/pnas.1604850113 – ident: ref222 doi: 10.1109/ISCA.2018.00060 – year: 2019 ident: ref106 article-title: Gate decorator: Global filter pruning method for accelerating deep convolutional neural networks publication-title: arXiv:1909.08174 – ident: ref329 doi: 10.1109/ISCA.2018.00012 – volume-title: DesignWare ARC EV Processors for Embedded Vision year: 2022 ident: ref306 – ident: ref171 doi: 10.1145/3287624.3287627 – year: 2017 ident: ref213 article-title: Adaptive neural networks for efficient inference publication-title: arXiv:1702.07811 – ident: ref355 doi: 10.1109/PDP2018.2018.00023 – volume-title: NVIDIA TensorRT year: 2022 ident: ref268 – year: 2019 ident: ref80 article-title: Efficient Winograd convolution via integer arithmetic publication-title: arXiv:1901.01965 – volume-title: Open Neural Network Exchange year: 2022 ident: ref260 – ident: ref320 doi: 10.1109/JSSC.2017.2778281 – ident: ref125 doi: 10.1109/DAC.2018.8465915 – ident: ref140 doi: 10.1007/978-3-319-46493-0_32 – ident: ref13 doi: 10.1109/TPDS.2020.3030548 – ident: ref191 doi: 10.1609/aaai.v34i04.5963 – ident: ref214 doi: 10.1109/ICDCS.2017.226 – ident: ref278 doi: 10.1145/3089801.3089804 – year: 2017 ident: ref99 article-title: Exploring the regularity of sparse structure in convolutional neural networks publication-title: arXiv:1705.08922 – ident: ref174 doi: 10.1145/2627369.2627613 – volume: 29 start-page: 2074 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref98 article-title: Learning structured sparsity in deep neural networks – ident: ref353 doi: 10.1016/j.pmcj.2017.07.014 – start-page: 2285 volume-title: Proc. ICML ident: ref149 article-title: Compressing neural networks with the hashing trick – start-page: 1269 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref202 article-title: Exploiting linear structure within convolutional networks for efficient evaluation – start-page: 3505 volume-title: Proc. 26th ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining ident: ref290 article-title: DeepSpeed – ident: ref325 doi: 10.23919/VLSIC.2019.8778056 – volume-title: Deploy Machine Learning Models on Mobile and IoT Devices year: 2022 ident: ref272 – ident: ref383 doi: 10.1109/TIFS.2015.2400395 – ident: ref400 doi: 10.1109/ISNE.2017.7968748 – ident: ref53 doi: 10.1109/CVPR.2016.90 – start-page: 963 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref135 article-title: Expectation backpropagation: Parameter-free training of multilayer neural networks with continuous or discrete weights – ident: ref304 doi: 10.1109/UCC48980.2020.00066 – year: 2014 ident: ref51 article-title: Very deep convolutional networks for large-scale image recognition publication-title: arXiv:1409.1556 – year: 2017 ident: ref152 article-title: Incremental network quantization: Towards lossless CNNs with low-precision weights publication-title: arXiv:1702.03044 – ident: ref197 doi: 10.1145/3089801.3089806 – volume: 32 start-page: 8026 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref257 article-title: PyTorch: An imperative style, high-performance deep learning library – ident: ref319 doi: 10.1145/3079856.3080246 – ident: ref208 doi: 10.1145/2906388.2906396 – year: 2022 ident: ref295 article-title: ZeroQuant: Efficient and affordable post-training quantization for large-scale transformers publication-title: arXiv:2206.01861 – volume-title: Products, Helping You Bring Local AI to Applications From Prototype to Production year: 2022 ident: ref312 – ident: ref220 doi: 10.1109/CVPR.2019.01147 – ident: ref347 doi: 10.1145/3161174 – year: 2018 ident: ref297 article-title: Horovod: Fast and easy distributed deep learning in TensorFlow publication-title: arXiv:1802.05799 – ident: ref237 doi: 10.1109/JETCAS.2019.2910232 – ident: ref341 doi: 10.1109/ASPDAC.2017.7858419 – volume-title: X-CUBE-AI, AI Expansion Pack for STM32CubeMX year: 2022 ident: ref266 – volume: 4107 volume-title: Proc. Advances in Neural Information Processing Systems ident: ref141 article-title: Binarized neural networks – volume: 15 start-page: 1929 issue: 1 year: 2014 ident: ref48 article-title: Dropout: A simple way to prevent neural networks from overfitting publication-title: J. Mach. Learn. Res. – year: 2017 ident: ref59 article-title: Learning transferable architectures for scalable image recognition publication-title: arXiv:1707.07012 – ident: ref233 doi: 10.1109/MICRO.2016.7783723 – year: 2016 ident: ref146 article-title: Trained ternary quantization publication-title: arXiv:1612.01064 – volume-title: DNNDK User Guide year: 2022 ident: ref267 – ident: ref389 doi: 10.3233/JIFS-169699 – ident: ref352 doi: 10.1109/ICCCN.2018.8487352 – ident: ref308 doi: 10.23919/VLSIC.2019.8778006 – year: 2014 ident: ref163 article-title: Compressing deep convolutional networks using vector quantization publication-title: arXiv:1412.6115 – ident: ref151 doi: 10.1109/I2MTC50364.2021.9459794 – ident: ref423 doi: 10.1109/MWC.006.2100699 – year: 2015 ident: ref157 article-title: Deep compression: Compressing deep neural networks with pruning, trained quantization and Huffman coding publication-title: arXiv:1510.00149 – ident: ref332 doi: 10.1109/EEEI.2014.7005803 – ident: ref130 doi: 10.1109/ISVLSI.2016.111 – volume-title: Qualcomm Neural Processing SDK for AI year: 2021 ident: ref270 – ident: ref415 doi: 10.1145/3240765.3240801 – ident: ref6 doi: 10.1109/MSP.2017.2765695 – ident: ref97 doi: 10.1145/3352460.3358269 – ident: ref309 doi: 10.23919/VLSIC.2019.8778193 – year: 2016 ident: ref205 article-title: Paying more attention to attention: Improving the performance of convolutional neural networks via attention transfer publication-title: arXiv:1612.03928 – ident: ref259 doi: 10.1145/3089801.3089805 – start-page: 1 volume-title: Proc. 56th Annu. Design Autom. Conf. ident: ref127 article-title: BitBlade: Area and energy-efficient precision-scalable neural network accelerator with bitwise summation – ident: ref322 doi: 10.1109/MM.2018.053631145 – ident: ref126 doi: 10.1109/ISCA.2018.00069 – volume-title: AI for the Edge year: 2022 ident: ref307 – start-page: 10118 volume-title: Proc. Int. Conf. Mach. Learn. ident: ref292 article-title: 1-bit Adam: Communication efficient large-scale training with Adam’s convergence speed – year: 2016 ident: ref137 article-title: BinaryNet: Training deep neural networks with weights and activations constrained to +1 or −1 publication-title: arXiv:1602.02830 – ident: ref406 doi: 10.1016/j.neucom.2017.09.046 – ident: ref81 doi: 10.1109/FCCM.2017.64 – ident: ref74 doi: 10.1109/CVPRW.2018.00215 – ident: ref21 doi: 10.1109/ICDCS.2017.94 – ident: ref9 doi: 10.1016/j.sysarc.2020.101887 – volume-title: Worldwide and U.S. IoT Cellular Connections Forecast, 2021–2025 year: 2022 ident: ref3 – ident: ref357 doi: 10.1609/aimag.v35i4.2513 – volume-title: CEVA Deep Neural Network (CDNN) year: 2022 ident: ref269 – ident: ref187 doi: 10.1609/aaai.v32i1.11601 – ident: ref150 doi: 10.1109/ICASSP.2017.7953288 – start-page: 50 volume-title: Proc. ACM Symp. Cloud Comput. ident: ref302 article-title: BigDL – ident: ref22 doi: 10.1145/3132211.3134456 – year: 2017 ident: ref70 article-title: MobileNets: Efficient convolutional neural networks for mobile vision applications publication-title: arXiv:1704.04861 – ident: ref95 doi: 10.1109/TC.2019.2914438 – year: 2021 ident: ref294 article-title: 1-bit Lamb: Communication efficient large-scale large-batch training with Lamb’s convergence speed publication-title: arXiv:2104.06069 – ident: ref396 doi: 10.1145/3154448.3154452 – ident: ref241 doi: 10.1109/ISCA.2018.00068 – ident: ref314 doi: 10.1145/3174243.3174253 – ident: ref386 doi: 10.1109/BTAS46853.2019.9185981 – volume-title: EdgeML: Machine learning for resource-constrained edge devices year: 2020 ident: ref30 – year: 2016 ident: ref61 article-title: Neural architecture search with reinforcement learning publication-title: arXiv:1611.01578 – ident: ref277 doi: 10.3390/electronics5040088 – ident: ref335 doi: 10.3390/s21092984 – ident: ref393 doi: 10.1145/3081333.3081359 – ident: ref419 doi: 10.1145/3079856.3080244 – ident: ref104 doi: 10.1109/ICPR.2018.8546129 – start-page: 551 volume-title: Proc. USENIX Annu. Tech. Conf. ident: ref291 article-title: ZeRO-Offload: Democratizing billion-scale model training – ident: ref49 doi: 10.1109/5.726791 – year: 2019 ident: ref155 article-title: A simplified fully quantized transformer for end-to-end speech recognition publication-title: arXiv:1911.03604 – ident: ref390 doi: 10.3390/s18113726 – year: 2018 ident: ref159 article-title: CMSIS-NN: Efficient neural network kernels for arm Cortex-M CPUs publication-title: arXiv:1801.06601 – ident: ref235 doi: 10.1109/MDAT.2017.2741463 – volume-title: Semi-supervised learning literature survey year: 2005 ident: ref40 – year: 2022 ident: ref102 article-title: Delta keyword transformer: Bringing transformers to the edge through dynamically pruned multi-head self-attention publication-title: arXiv:2204.03479 – ident: ref200 doi: 10.1145/2994551.2994564 – year: 2016 ident: ref35 article-title: An overview of gradient descent optimization algorithms publication-title: arXiv:1609.04747 – ident: ref398 doi: 10.1145/3131895 – ident: ref413 doi: 10.1145/3242897 – ident: ref422 doi: 10.2200/s00832ed1v01y201802aim037 – ident: ref342 doi: 10.1109/DAC.2018.8465793 – ident: ref361 doi: 10.1145/3494834.3500240 – ident: ref240 doi: 10.1109/ISCA.2018.00061 – volume: 5 start-page: 532 year: 2020 ident: ref33 article-title: Cross-validation publication-title: Encyclopedia Database Syst. – ident: ref68 doi: 10.18653/v1/2020.acl-main.686 – ident: ref103 doi: 10.1109/ISQED54688.2022.9806197 – ident: ref287 doi: 10.1145/2996864 – start-page: 1 volume-title: Proc. 15th ACM Conf. Embedded Netw. Sensor Syst. ident: ref158 article-title: DeepIoT: Compressing deep neural network structures for sensing systems with a compressor-critic framework – volume-title: Jetson TX2 Module year: 2022 ident: ref253 – ident: ref378 doi: 10.1145/3229556.3229562 – ident: ref424 doi: 10.1109/ACCESS.2018.2870052 – ident: ref239 doi: 10.1145/3297858.3304028 – ident: ref111 doi: 10.1109/FCCM.2017.47 – year: 2022 ident: ref156 article-title: I-ViT: Integer-only quantization for efficient vision transformer inference publication-title: arXiv:2207.01405 – year: 2018 ident: ref113 article-title: Training deep neural networks with 8-bit floating point numbers publication-title: arXiv:1812.08011 – ident: ref262 doi: 10.1145/3373376.3378534 – ident: ref344 doi: 10.1145/3287624.3287715 – start-page: 236 volume-title: Proc. Int. Symp. Comput. Archit. ident: ref345 article-title: Sparse ReRAM engine: Joint exploration of activation and weight sparsity in compressed neural networks – ident: ref182 doi: 10.48550/arXiv.1503.02531 – ident: ref196 doi: 10.1109/CVPR.2018.00171 – ident: ref372 doi: 10.1007/978-3-030-27562-4_6 – ident: ref243 doi: 10.1145/3007787.3001163 – year: 2016 ident: ref145 article-title: Ternary weight networks publication-title: arXiv:1605.04711 – ident: ref420 doi: 10.1109/HPCA.2019.00027 – ident: ref199 doi: 10.1109/EMC2-NIPS53020.2019.00013 – ident: ref348 doi: 10.1145/2820975.2820980 – start-page: 1117 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref136 article-title: Backpropagation for energy-efficient neuromorphic computing – year: 2017 ident: ref172 article-title: ApproxDBN: Approximate computing for discriminative deep belief networks publication-title: arXiv:1704.03993 – ident: ref281 doi: 10.1145/2935643.2935650 – start-page: 1 volume-title: Proc. Deep Learn. Unsupervised Feature Learn. Workshop ident: ref120 article-title: Improving the speed of neural networks on CPUs – ident: ref217 doi: 10.1007/978-3-030-01261-8_25 – ident: ref373 doi: 10.1109/SmartWorld-UIC-ATC-SCALCOM-IOP-SCI.2019.00194 – ident: ref164 doi: 10.1145/3289602.3293902 – start-page: 701 volume-title: Proc. Design, Autom. Test Eur. Conf. Exhib. (DATE) ident: ref170 article-title: ApproxANN: An approximate computing framework for artificial neural network – year: 2015 ident: ref175 article-title: Compression of deep convolutional neural networks for fast and low power mobile applications publication-title: arXiv:1511.06530 – ident: ref216 doi: 10.1109/ICCD.2017.49 – ident: ref376 doi: 10.1109/TNET.2020.3042320 – ident: ref19 doi: 10.1109/JPROC.2019.2921977 – ident: ref244 doi: 10.1145/3140659.3080254 – volume: 25 start-page: 1097 volume-title: Proc. Adv. Neural Inf. Process. Syst. (NIPS) ident: ref50 article-title: ImageNet classification with deep convolutional neural networks – ident: ref249 doi: 10.1109/TC.2019.2924215 – start-page: 1822 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref92 article-title: Architectural complexity measures of recurrent neural networks – ident: ref231 doi: 10.1109/ICTAI.2019.00197 – ident: ref256 doi: 10.1145/2647868.2654889 – ident: ref336 doi: 10.1145/3007787.3001140 – volume-title: AutoML year: 2022 ident: ref62 – ident: ref129 doi: 10.1145/3352460.3358295 – year: 2019 ident: ref301 article-title: TF-replicator: Distributed machine learning for researchers publication-title: arXiv:1902.00465 – ident: ref55 doi: 10.1609/aaai.v31i1.11231 – ident: ref131 doi: 10.23919/VLSIC.2017.8008533 – start-page: 2407 volume-title: Proc. 24th ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining ident: ref366 article-title: Not just privacy – ident: ref5 doi: 10.3390/electronics8030292 – volume-title: FairScale: A general purpose modular PyTorch library for high performance and large scale training year: 2021 ident: ref299 – ident: ref226 doi: 10.18653/v1/2020.emnlp-demos.6 – year: 2017 ident: ref90 article-title: Factorization tricks for LSTM networks publication-title: arXiv:1703.10722 – ident: ref401 doi: 10.1145/3299710.3211336 – ident: ref72 doi: 10.1109/CVPR.2018.00474 – ident: ref236 doi: 10.1109/ICPR.2018.8545462 – year: 2020 ident: ref298 article-title: Benchmark tests of convolutional neural network and graph convolutional network on HorovodRunner enabled spark clusters publication-title: arXiv:2005.05510 – start-page: 561 volume-title: Proc. 13th USENIX Symp. Operating Syst. Design Implement. (OSDI) ident: ref303 article-title: Ray: A distributed framework for emerging AI applications – ident: ref26 doi: 10.1145/3267809.3267828 – ident: ref177 doi: 10.1145/3093336.3037746 – ident: ref387 doi: 10.1109/AIPR50011.2020.9425332 – ident: ref343 doi: 10.1109/ACCESS.2018.2874823 – ident: ref169 doi: 10.1109/CASES.2015.7324548 – volume-title: Machine Learning: A Probabilistic Perspective year: 2012 ident: ref32 – year: 2014 ident: ref44 article-title: Learning activation functions to improve deep neural networks publication-title: arXiv:1412.6830 – ident: ref283 doi: 10.1109/CVPR.2016.91 – ident: ref183 doi: 10.1609/aaai.v33i01.33013779 – start-page: 145 volume-title: Proc. Design, Autom. Test Eur. Conf. Exhib. (DATE) ident: ref166 article-title: Multiplier-less artificial neurons exploiting error resiliency for energy-efficient neural computing – ident: ref248 doi: 10.1109/TNNLS.2018.2852335 – ident: ref186 doi: 10.1609/aaai.v32i1.11876 – ident: ref123 doi: 10.1109/LCA.2016.2597140 – year: 2019 ident: ref77 article-title: ANTNets: Mobile convolutional neural networks for resource efficient image classification publication-title: arXiv:1904.03775 – ident: ref421 doi: 10.1145/3472291 – year: 2015 ident: ref93 article-title: Grid long short-term memory publication-title: arXiv:1507.01526 – year: 2018 ident: ref368 article-title: PipeDream: Fast and efficient pipeline parallel DNN training publication-title: arXiv:1806.03377 – ident: ref411 doi: 10.1111/1754-9485.13261 – start-page: 265 volume-title: Proc. 12th USENIX Symp. Operating Syst. Design Implement. ident: ref255 article-title: TensorFlow: A system for large-scale machine learning – start-page: 1 volume-title: Proc. USENIX Workshop Hot Topics Edge Comput. ident: ref370 article-title: MODI: Mobile deep inference made efficient by edge computing – ident: ref337 doi: 10.1038/s41586-018-0180-5 – ident: ref232 doi: 10.1109/ISSCC.2019.8662302 – ident: ref117 doi: 10.1109/ICASSP.2015.7178146 – year: 2018 ident: ref69 article-title: DARTS: Differentiable architecture search publication-title: arXiv:1806.09055 – year: 2017 ident: ref184 article-title: Like what you like: Knowledge distill via neuron selectivity transfer publication-title: arXiv:1707.01219 – year: 2017 ident: ref198 article-title: SquishedNets: Squishing SqueezeNet further for edge device scenarios via deep evolutionary synthesis publication-title: arXiv:1711.07459 – ident: ref87 doi: 10.1007/s11633-016-1006-2 – year: 2016 ident: ref142 article-title: Quantized neural networks: Training neural networks with low precision weights and activations publication-title: arXiv:1609.07061 – ident: ref83 doi: 10.1145/3289602.3293905 – ident: ref38 doi: 10.1007/978-0-387-84858-7_14 – ident: ref100 doi: 10.1109/WACV.2018.00083 – ident: ref408 doi: 10.1109/ICCCBDA51879.2021.9442600 – ident: ref67 doi: 10.1109/CVPR.2019.00293 – year: 2015 ident: ref133 article-title: Neural networks with few multiplications publication-title: arXiv:1510.03009 – year: 2020 ident: ref227 article-title: SpAtten: Efficient sparse attention architecture with cascade token and head pruning publication-title: arXiv:2012.09852 – ident: ref369 doi: 10.21437/Interspeech.2021-1816 – volume-title: AutoKeras year: 2022 ident: ref63 – start-page: 1 volume-title: Proc. 11th IEEE/ACM/IFIP Int. Conf. Hardw./Softw. Codesign Syst. Synth. ident: ref211 article-title: Runtime configurable deep neural networks for energy-accuracy trade-off – year: 2020 ident: ref194 article-title: Flexible dataset distillation: Learn labels instead of images publication-title: arXiv:2006.08572 – ident: ref168 doi: 10.1109/ISLPED.2017.8009173 – ident: ref246 doi: 10.1145/3352460.3358291 – start-page: 231 volume-title: Pro Android Apps Performance Optimization year: 2012 ident: ref279 article-title: RenderScript doi: 10.1007/978-1-4302-4000-6_9 – ident: ref261 doi: 10.1609/aaai.v34i04.5954 – ident: ref212 doi: 10.1109/ICPR.2016.7900006 – ident: ref371 doi: 10.23919/DATE.2017.7927211 – ident: ref12 doi: 10.1007/s00521-018-3761-1 – ident: ref275 doi: 10.1609/aaai.v35i2.16179 – ident: ref416 doi: 10.1109/FCCM.2019.00033 – ident: ref310 doi: 10.1109/VLSIC.2018.8502404 – ident: ref403 doi: 10.1109/ISCAS51556.2021.9401083 – ident: ref2 doi: 10.1109/JPROC.2017.2761740 – ident: ref201 doi: 10.1145/3109761.3109804 – ident: ref409 doi: 10.1109/TIM.2020.3018831 – ident: ref273 doi: 10.1086/719650 – ident: ref39 doi: 10.1007/s10994-019-05855-6 – ident: ref1 doi: 10.7551/mitpress/12832.003.0015 – start-page: 1 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref300 article-title: Mesh-TensorFlow: Deep learning for supercomputers – year: 2017 ident: ref24 article-title: DeepCache: Principled cache for mobile deep vision publication-title: arXiv:1712.01670 – ident: ref333 doi: 10.23919/DATE48585.2020.9116529 – ident: ref350 doi: 10.1109/SEC.2016.38 – ident: ref234 doi: 10.1109/LCA.2020.2979965 – ident: ref349 doi: 10.1145/3318216.3363316 – ident: ref190 doi: 10.1109/CVPR.2018.00454 – ident: ref57 doi: 10.1109/CVPR.2017.243 – year: 2017 ident: ref221 article-title: Rethinking atrous convolution for semantic image segmentation publication-title: arXiv:1706.05587 – ident: ref65 doi: 10.1007/978-3-030-01249-6_18 – start-page: 5998 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref45 article-title: Attention is all you need – ident: ref124 doi: 10.1109/ISSCC.2018.8310262 – ident: ref285 doi: 10.4108/eai.30-11-2016.2267463 – ident: ref250 doi: 10.1109/JSSC.2016.2616357 – ident: ref375 doi: 10.1109/TCAD.2018.2858384 – ident: ref42 doi: 10.1613/jair.301 – ident: ref252 doi: 10.3390/en15207495 – ident: ref11 doi: 10.1109/ACCESS.2018.2890150 – ident: ref79 doi: 10.1109/CVPR.2016.435 – ident: ref223 doi: 10.1109/CVPR42600.2020.01464 – ident: ref313 doi: 10.1145/3174243.3174261 – ident: ref66 doi: 10.1007/978-3-030-01246-5_2 – year: 2017 ident: ref334 article-title: Hello edge: Keyword spotting on microcontrollers publication-title: arXiv:1711.07128 – ident: ref119 doi: 10.1109/VLSIC.2016.7573525 – year: 2020 ident: ref46 article-title: Language models are few-shot learners publication-title: arXiv:2005.14165 – start-page: 2888 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref204 article-title: Moonshine: Distilling with cheap convolutions – start-page: 916 volume-title: Are very deep neural networks feasible on mobile devices? year: 2016 ident: ref282 – ident: ref289 doi: 10.1145/3296957.3173176 – ident: ref388 doi: 10.1109/PERCOMW.2016.7457169 – year: 2022 ident: ref71 article-title: EdgeNeXt: Efficiently amalgamated CNN-transformer architecture for mobile vision applications publication-title: arXiv:2206.10589 – ident: ref122 doi: 10.1109/EMC2.2018.00012 – ident: ref160 doi: 10.1145/3020078.3021745 – ident: ref288 doi: 10.1109/CVPRW.2011.5981829 – ident: ref203 doi: 10.5244/C.28.88 – year: 2016 ident: ref363 article-title: Federated learning: Strategies for improving communication efficiency publication-title: arXiv:1610.05492 – ident: ref280 doi: 10.1109/LES.2018.2815954 – start-page: 1273 volume-title: Artificial Intelligence and Statistics year: 2017 ident: ref359 article-title: Communication-efficient learning of deep networks from decentralized data – ident: ref346 doi: 10.1109/MPRV.2017.2940968 – ident: ref52 doi: 10.4324/9781410605337-29 – year: 2018 ident: ref418 article-title: ProxylessNAS: Direct neural architecture search on target task and hardware publication-title: arXiv:1812.00332 – ident: ref364 doi: 10.1109/LCOMM.2019.2921755 – year: 2020 ident: ref274 article-title: CoCoPIE: Making mobile AI sweet as PIE—Compression-compilation co-design goes a long way publication-title: arXiv:2003.06700 – ident: ref188 doi: 10.18653/v1/2020.findings-emnlp.372 – ident: ref110 doi: 10.1145/3020078.3021740 – ident: ref109 doi: 10.1016/j.micpro.2020.102991 – ident: ref58 doi: 10.1109/CVPR.2018.00745 – ident: ref321 doi: 10.23919/VLSIC.2017.8008534 – ident: ref397 doi: 10.1145/2750858.2804262 – ident: ref128 doi: 10.1109/FPL.2018.00035 – volume: 48 start-page: 2849 volume-title: Proc. 33rd Int. Conf. Int. Conf. Mach. Learn. ident: ref115 article-title: Fixed point quantization of deep convolutional networks – ident: ref167 doi: 10.1109/JETCAS.2018.2835809 – ident: ref316 doi: 10.1109/TNNLS.2018.2844093 – ident: ref179 doi: 10.1145/3287624.3287714 – ident: ref238 doi: 10.1109/MICRO.2018.00024 – volume-title: Deep Learning HDL Toolbox year: 2022 ident: ref265 – ident: ref144 doi: 10.1109/CVPR.2017.574 – ident: ref276 doi: 10.1145/2964284.2973801 – ident: ref75 doi: 10.1109/CVPR.2018.00716 – year: 2022 ident: ref293 article-title: Maximizing communication efficiency for large-scale training via 0/1 Adam publication-title: arXiv:2202.06009 – ident: ref10 doi: 10.1109/JPROC.2020.2976475 – ident: ref374 doi: 10.1109/PADSW.2018.8645013 – ident: ref56 doi: 10.1109/CVPR.2017.195 – ident: ref417 doi: 10.1145/3289602.3293910 – ident: ref414 doi: 10.1109/MICRO.2016.7783720 – ident: ref28 doi: 10.1109/INFOCOM.2018.8486403 – ident: ref404 doi: 10.3390/electronics8111321 – ident: ref360 doi: 10.2200/s00960ed2v01y201910aim043 – ident: ref242 doi: 10.1109/MICRO.2018.00011 – ident: ref139 doi: 10.1007/978-3-030-01267-0_44 – ident: ref385 doi: 10.1007/978-3-319-97909-0_46 – ident: ref219 doi: 10.1109/EMBC.2017.8036767 – start-page: 3882 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref88 article-title: Phased LSTM: Accelerating recurrent network training for long or event-based sequences – year: 2016 ident: ref134 article-title: Binarized neural networks: Training deep neural networks with weights and activations constrained to +1 or −1 publication-title: arXiv:1602.02830 – ident: ref192 doi: 10.1109/CVPR42600.2020.00396 – volume: 64 start-page: 62 issue: 6 year: 2021 ident: ref263 article-title: CoCoPIE publication-title: Commun. ACM doi: 10.1145/3418297 – ident: ref85 doi: 10.1109/ICASSP.2016.7472657 – ident: ref379 doi: 10.1109/IPDPS53621.2022.00110 – ident: ref181 doi: 10.1145/1150402.1150464 – ident: ref215 doi: 10.1109/TII.2018.2842821 – ident: ref224 doi: 10.48550/arXiv.1810.04805 – ident: ref317 doi: 10.1109/FPL.2018.00075 – year: 2018 ident: ref86 article-title: The unreasonable effectiveness of the forget gate publication-title: arXiv:1804.04849 – ident: ref4 doi: 10.1145/3234150 – ident: ref34 doi: 10.1016/j.patcog.2015.03.009 – ident: ref391 doi: 10.1145/3212725.3212728 – ident: ref354 doi: 10.1109/ICPADS.2017.00069 – ident: ref358 doi: 10.1109/CHASE.2017.115 – ident: ref218 doi: 10.1109/ISSCC.2017.7870353 – ident: ref380 doi: 10.1109/INFOCOM48880.2022.9796896 – ident: ref162 doi: 10.1145/3081333.3081360 – ident: ref14 doi: 10.1109/ICCVW.2019.00447 – ident: ref108 doi: 10.1109/CVPR.2017.643 – ident: ref78 doi: 10.1137/1.9781611970364 – ident: ref323 doi: 10.1109/TVLSI.2018.2797600 – ident: ref402 doi: 10.1145/3240765.3243473 – ident: ref76 doi: 10.1109/CVPR.2018.00291 – ident: ref121 doi: 10.1109/ICACSIS.2017.8355051 – ident: ref36 doi: 10.1007/978-3-540-75171-7_2 – ident: ref96 doi: 10.1145/3289602.3293898 – start-page: 3123 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref132 article-title: BinaryConnect: Training deep neural networks with binary weights during propagations – ident: ref315 doi: 10.1109/TCAD.2017.2705069 – year: 2014 ident: ref89 article-title: Long short-term memory based recurrent neural network architectures for large vocabulary speech recognition publication-title: arXiv:1402.1128 – ident: ref161 doi: 10.1109/TMC.2020.2999956 – start-page: 601 volume-title: Proc. Adv. Neural Inf. Process. Syst. ident: ref154 article-title: HitNet: Hybrid ternary recurrent neural network
SSID	ssj0003563
Score	2.7179554
SecondaryResourceType	review_article
Snippet	Successful integration of deep neural networks (DNNs) or deep learning (DL) has resulted in breakthroughs in many areas. However, deploying these highly...
SourceID	proquest crossref ieee
SourceType	Aggregation Database Enrichment Source Index Database Publisher
StartPage	1
SubjectTerms	Algorithms Algorithm–hardware codesign Artificial intelligence artificial intelligence (AI) artificial intelligence on edge (edge-AI) Artificial neural networks Cloud computing Co-design Computer architecture Cyber-physical systems Data transmission Deep learning deep learning (DL) Design optimization Hardware Image edge detection Inference Internet of Things Iterative methods Machine learning Memory devices model compression Network latency neural accelerator Optimization Optimization techniques Real-time systems State-of-the-art reviews Training
Title	Efficient Acceleration of Deep Learning Inference on Resource-Constrained Edge Devices: A Review
URI	https://ieeexplore.ieee.org/document/9985008 https://www.proquest.com/docview/2763809889
Volume	111
WOSCitedRecordID	wos000899996300001&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/IET Electronic Library customDbUrl: eissn: 1558-2256 dateEnd: 99991231 omitProxy: false ssIdentifier: ssj0003563 issn: 0018-9219 databaseCode: RIE dateStart: 19630101 isFulltext: true titleUrlDefault: https://ieeexplore.ieee.org/ providerName: IEEE
link	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1JSwMxFA61eNCDWxWrVXLwptPO2mS8ldqiIrWIQm9jlpciSFu6-Pt9ycwURRG8DUwehPkmb0nyfY-QC6MTDAOCe0msmRdL4XtCSIGAYPYuZOJLcEThBzYY8NEoHVbI1ZoLAwDu8hk07aM7y9dTtbJbZS0sDRLH7N1grJ1ztdZeN0qKrmkBLmBchiVBxk9b98Onxy6WgmHYjCxvlAffgpDrqvLDFbv40t_938z2yE6RR9JODvw-qcDkgGx_UReskdeek4dAQ9pRCsNLDjadGnoDMKOFtOqY3pWkP4pvy_18z7bydA0kQNOeHgMaOa9yTTs0P1E4JC_93nP31isaKngK87Klh9Hc10zoODRY-br80GCAV9KYUIS8LTkTMSQmCgJlmBYQpsLoSHIrCiQVj45IdTKdwDGhiKTyMbkTmI7E4Le5ARlzK3YDDEs4XSdB-YUzVaiN2zm_Z67q8NPMoZJZVLIClTq5XNvMcq2NP0fXLA7rkQUEddIogcyK5bjIQvSi3E85T09-tzolW7aPfL630iDV5XwFZ2RTfSzfFvNz96d9AhSrz60
linkProvider	IEEE
linkToHtml	http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1JSwMxFH6ICurBXaxrDt50dNZOxlvRikutIgrexiwvRZBWbPX3-5LJFEURvA1MHoT5Jm9J8n0PYM_ojMKA4EGW6jxIpQgDIaQgQCh7FzILJTqicCfvdvnjY3E7AQdjLgwiustneGgf3Vm-Hqh3u1V2RKVB5pi9U7Zzlmdrjf1ukvm-aREtYVqINUUmLI4ub-9uTqgYjOPDxDJHefQtDLm-Kj-csYswZwv_m9sizPtMkrUq6JdgAvvLMPdFX3AFntpOIIIMWUspCjAV3Gxg2CniK_Piqj12UdP-GL2td_QD28zTtZBAzdq6h2Tk_Moxa7HqTGEVHs7a9yfngW-pECjKzEYBxfNQ50KnsaHa12WIhkK8ksbEIuZNyXORYmaSKFIm1wLjQhidSG5lgaTiyRpM9gd9XAdGWKqQ0jtBCUmKYZMblCm3cjeYUxGnGxDVX7hUXm_czvmldHVHWJQOldKiUnpUGrA_tnmt1Db-HL1icRiP9BA0YKsGsvQLcljG5Ed5WHBebPxutQsz5_fXnbJz0b3ahFnbVb7aadmCydHbO27DtPoYPQ_fdtxf9wkSU9L2
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=Efficient+Acceleration+of+Deep+Learning+Inference+on+Resource-Constrained+Edge+Devices%3A+A+Review&rft.jtitle=Proceedings+of+the+IEEE&rft.au=Shuvo%2C+Md.+Maruf+Hossain&rft.au=Islam%2C+Syed+Kamrul&rft.au=Cheng%2C+Jianlin&rft.au=Morshed%2C+Bashir+I.&rft.date=2023-01-01&rft.pub=IEEE&rft.issn=0018-9219&rft.spage=1&rft.epage=50&rft_id=info:doi/10.1109%2FJPROC.2022.3226481&rft.externalDocID=9985008
thumbnail_l	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0018-9219&client=summon
thumbnail_m	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0018-9219&client=summon
thumbnail_s	http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0018-9219&client=summon