Sustainable Edge Intelligence Through Energy-Aware Early Exiting

Deep learning (DL) models have emerged as a promising solution for the Internet of Things (IoT). However, due to their computational complexity, DL models consume significant amounts of energy, which can rapidly drain the battery and compromise the performance of IoT devices. For sustainable operati...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:arXiv.org 2023-07
Hauptverfasser: Bullo, Marcello, Jardak, Seifallah, Carnelli, Pietro, Gündüz, Deniz
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Bullo, Marcello
Jardak, Seifallah
Carnelli, Pietro
Gündüz, Deniz
description Deep learning (DL) models have emerged as a promising solution for the Internet of Things (IoT). However, due to their computational complexity, DL models consume significant amounts of energy, which can rapidly drain the battery and compromise the performance of IoT devices. For sustainable operation, we consider an edge device with a rechargeable battery and energy harvesting (EH) capabilities. In addition to the stochastic nature of the ambient energy source, the harvesting rate is often insufficient to meet the inference energy requirements, leading to drastic performance degradation in energy-agnostic devices. To mitigate this problem, we propose energy-adaptive dynamic early exiting (EE) to enable efficient and accurate inference in an EH edge intelligence system. Our approach derives an energy-aware EE policy that determines the optimal amount of computational processing on a per-sample basis. The proposed policy balances the energy consumption to match the limited incoming energy and achieves continuous availability. Numerical results show that accuracy and service rate are improved up to 25% and 35%, respectively, in comparison with an energy-agnostic policy.
doi_str_mv 10.48550/arxiv.2305.14094
format Article
fullrecord <record><control><sourceid>proquest_arxiv</sourceid><recordid>TN_cdi_arxiv_primary_2305_14094</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2818535813</sourcerecordid><originalsourceid>FETCH-LOGICAL-a954-5001a8c62738fb85b99721591050de000ce94a82d061a57361b91885eb7a99733</originalsourceid><addsrcrecordid>eNotj81Og0AURicmJja1D-BKEtfUOz8XZnY2DWqTJi5kTy4wpdMg1AG0vL20dfVtTr6cw9gDh6XSiPBM_uR-lkICLrkCo27YTEjJQ62EuGOLrjsAgIhigShn7OVz6HpyDeW1DZKyssGm6W1du8o2hQ3SvW-Hah8kjfXVGK5-yU8Y-XoMkpPrXVPds9sd1Z1d_O-cpa9Jun4Ptx9vm_VqG5JBFSIAJ11EIpZ6l2vMjYkFR8MBobSTUGGNIi1KiDhhLCOeG6412jymCZVyzh6vt5e87OjdF_kxO2dml8yJeLoSR99-D7brs0M7-GZyyoTmGiVqLuUfGRhS7w</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2818535813</pqid></control><display><type>article</type><title>Sustainable Edge Intelligence Through Energy-Aware Early Exiting</title><source>arXiv.org</source><source>Free E- Journals</source><creator>Bullo, Marcello ; Jardak, Seifallah ; Carnelli, Pietro ; Gündüz, Deniz</creator><creatorcontrib>Bullo, Marcello ; Jardak, Seifallah ; Carnelli, Pietro ; Gündüz, Deniz</creatorcontrib><description>Deep learning (DL) models have emerged as a promising solution for the Internet of Things (IoT). However, due to their computational complexity, DL models consume significant amounts of energy, which can rapidly drain the battery and compromise the performance of IoT devices. For sustainable operation, we consider an edge device with a rechargeable battery and energy harvesting (EH) capabilities. In addition to the stochastic nature of the ambient energy source, the harvesting rate is often insufficient to meet the inference energy requirements, leading to drastic performance degradation in energy-agnostic devices. To mitigate this problem, we propose energy-adaptive dynamic early exiting (EE) to enable efficient and accurate inference in an EH edge intelligence system. Our approach derives an energy-aware EE policy that determines the optimal amount of computational processing on a per-sample basis. The proposed policy balances the energy consumption to match the limited incoming energy and achieves continuous availability. Numerical results show that accuracy and service rate are improved up to 25% and 35%, respectively, in comparison with an energy-agnostic policy.</description><identifier>EISSN: 2331-8422</identifier><identifier>DOI: 10.48550/arxiv.2305.14094</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Batteries ; Computer Science - Artificial Intelligence ; Computer Science - Systems and Control ; Energy consumption ; Energy harvesting ; Energy management ; Energy requirements ; Inference ; Intelligence ; Internet of Things ; Microbalances ; Performance degradation ; Rechargeable batteries ; Statistics - Machine Learning</subject><ispartof>arXiv.org, 2023-07</ispartof><rights>2023. This work is published under http://arxiv.org/licenses/nonexclusive-distrib/1.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,780,881,27902</link.rule.ids><backlink>$$Uhttps://doi.org/10.1109/MLSP55844.2023.10285956$$DView published paper (Access to full text may be restricted)$$Hfree_for_read</backlink><backlink>$$Uhttps://doi.org/10.48550/arXiv.2305.14094$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Bullo, Marcello</creatorcontrib><creatorcontrib>Jardak, Seifallah</creatorcontrib><creatorcontrib>Carnelli, Pietro</creatorcontrib><creatorcontrib>Gündüz, Deniz</creatorcontrib><title>Sustainable Edge Intelligence Through Energy-Aware Early Exiting</title><title>arXiv.org</title><description>Deep learning (DL) models have emerged as a promising solution for the Internet of Things (IoT). However, due to their computational complexity, DL models consume significant amounts of energy, which can rapidly drain the battery and compromise the performance of IoT devices. For sustainable operation, we consider an edge device with a rechargeable battery and energy harvesting (EH) capabilities. In addition to the stochastic nature of the ambient energy source, the harvesting rate is often insufficient to meet the inference energy requirements, leading to drastic performance degradation in energy-agnostic devices. To mitigate this problem, we propose energy-adaptive dynamic early exiting (EE) to enable efficient and accurate inference in an EH edge intelligence system. Our approach derives an energy-aware EE policy that determines the optimal amount of computational processing on a per-sample basis. The proposed policy balances the energy consumption to match the limited incoming energy and achieves continuous availability. Numerical results show that accuracy and service rate are improved up to 25% and 35%, respectively, in comparison with an energy-agnostic policy.</description><subject>Batteries</subject><subject>Computer Science - Artificial Intelligence</subject><subject>Computer Science - Systems and Control</subject><subject>Energy consumption</subject><subject>Energy harvesting</subject><subject>Energy management</subject><subject>Energy requirements</subject><subject>Inference</subject><subject>Intelligence</subject><subject>Internet of Things</subject><subject>Microbalances</subject><subject>Performance degradation</subject><subject>Rechargeable batteries</subject><subject>Statistics - Machine Learning</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>BENPR</sourceid><sourceid>GOX</sourceid><recordid>eNotj81Og0AURicmJja1D-BKEtfUOz8XZnY2DWqTJi5kTy4wpdMg1AG0vL20dfVtTr6cw9gDh6XSiPBM_uR-lkICLrkCo27YTEjJQ62EuGOLrjsAgIhigShn7OVz6HpyDeW1DZKyssGm6W1du8o2hQ3SvW-Hah8kjfXVGK5-yU8Y-XoMkpPrXVPds9sd1Z1d_O-cpa9Jun4Ptx9vm_VqG5JBFSIAJ11EIpZ6l2vMjYkFR8MBobSTUGGNIi1KiDhhLCOeG6412jymCZVyzh6vt5e87OjdF_kxO2dml8yJeLoSR99-D7brs0M7-GZyyoTmGiVqLuUfGRhS7w</recordid><startdate>20230716</startdate><enddate>20230716</enddate><creator>Bullo, Marcello</creator><creator>Jardak, Seifallah</creator><creator>Carnelli, Pietro</creator><creator>Gündüz, Deniz</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20230716</creationdate><title>Sustainable Edge Intelligence Through Energy-Aware Early Exiting</title><author>Bullo, Marcello ; Jardak, Seifallah ; Carnelli, Pietro ; Gündüz, Deniz</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a954-5001a8c62738fb85b99721591050de000ce94a82d061a57361b91885eb7a99733</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Batteries</topic><topic>Computer Science - Artificial Intelligence</topic><topic>Computer Science - Systems and Control</topic><topic>Energy consumption</topic><topic>Energy harvesting</topic><topic>Energy management</topic><topic>Energy requirements</topic><topic>Inference</topic><topic>Intelligence</topic><topic>Internet of Things</topic><topic>Microbalances</topic><topic>Performance degradation</topic><topic>Rechargeable batteries</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Bullo, Marcello</creatorcontrib><creatorcontrib>Jardak, Seifallah</creatorcontrib><creatorcontrib>Carnelli, Pietro</creatorcontrib><creatorcontrib>Gündüz, Deniz</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni Edition)</collection><collection>ProQuest Central UK/Ireland</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central Korea</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection><jtitle>arXiv.org</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Bullo, Marcello</au><au>Jardak, Seifallah</au><au>Carnelli, Pietro</au><au>Gündüz, Deniz</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Sustainable Edge Intelligence Through Energy-Aware Early Exiting</atitle><jtitle>arXiv.org</jtitle><date>2023-07-16</date><risdate>2023</risdate><eissn>2331-8422</eissn><abstract>Deep learning (DL) models have emerged as a promising solution for the Internet of Things (IoT). However, due to their computational complexity, DL models consume significant amounts of energy, which can rapidly drain the battery and compromise the performance of IoT devices. For sustainable operation, we consider an edge device with a rechargeable battery and energy harvesting (EH) capabilities. In addition to the stochastic nature of the ambient energy source, the harvesting rate is often insufficient to meet the inference energy requirements, leading to drastic performance degradation in energy-agnostic devices. To mitigate this problem, we propose energy-adaptive dynamic early exiting (EE) to enable efficient and accurate inference in an EH edge intelligence system. Our approach derives an energy-aware EE policy that determines the optimal amount of computational processing on a per-sample basis. The proposed policy balances the energy consumption to match the limited incoming energy and achieves continuous availability. Numerical results show that accuracy and service rate are improved up to 25% and 35%, respectively, in comparison with an energy-agnostic policy.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><doi>10.48550/arxiv.2305.14094</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2023-07
issn 2331-8422
language eng
recordid cdi_arxiv_primary_2305_14094
source arXiv.org; Free E- Journals
subjects Batteries
Computer Science - Artificial Intelligence
Computer Science - Systems and Control
Energy consumption
Energy harvesting
Energy management
Energy requirements
Inference
Intelligence
Internet of Things
Microbalances
Performance degradation
Rechargeable batteries
Statistics - Machine Learning
title Sustainable Edge Intelligence Through Energy-Aware Early Exiting
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-30T08%3A07%3A21IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_arxiv&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Sustainable%20Edge%20Intelligence%20Through%20Energy-Aware%20Early%20Exiting&rft.jtitle=arXiv.org&rft.au=Bullo,%20Marcello&rft.date=2023-07-16&rft.eissn=2331-8422&rft_id=info:doi/10.48550/arxiv.2305.14094&rft_dat=%3Cproquest_arxiv%3E2818535813%3C/proquest_arxiv%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2818535813&rft_id=info:pmid/&rfr_iscdi=true