{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,7]],"date-time":"2026-08-07T18:40:42Z","timestamp":1786128042215,"version":"3.56.0"},"reference-count":62,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"10","license":[{"start":{"date-parts":[[2025,10,1]],"date-time":"2025-10-01T00:00:00Z","timestamp":1759276800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,10,1]],"date-time":"2025-10-01T00:00:00Z","timestamp":1759276800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,1]],"date-time":"2025-10-01T00:00:00Z","timestamp":1759276800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"PNRR-PE-AI FAIR project"},{"name":"NextGeneration EU Program"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Neural Netw. Learning Syst."],"published-print":{"date-parts":[[2025,10]]},"DOI":"10.1109\/tnnls.2025.3588558","type":"journal-article","created":{"date-parts":[[2025,7,21]],"date-time":"2025-07-21T18:09:25Z","timestamp":1753121365000},"page":"19342-19355","source":"Crossref","is-referenced-by-count":5,"title":["NACHOS: Neural Architecture Search for Hardware-Constrained Early-Exit Neural Networks"],"prefix":"10.1109","volume":"36","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-9734-670X","authenticated-orcid":false,"given":"Matteo","family":"Gambella","sequence":"first","affiliation":[{"name":"Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB), Politecnico di Milano, Milan, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3236-3941","authenticated-orcid":false,"given":"Jary","family":"Pomponi","sequence":"additional","affiliation":[{"name":"Dipartimento di Ingegneria dell'Informazione, Elettronica e Telecomunicazioni (DIET), Sapienza Universitàdi Roma, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0881-8344","authenticated-orcid":false,"given":"Simone","family":"Scardapane","sequence":"additional","affiliation":[{"name":"Dipartimento di Ingegneria dell'Informazione, Elettronica e Telecomunicazioni (DIET), Sapienza Universitàdi Roma, Rome, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7828-7687","authenticated-orcid":false,"given":"Manuel","family":"Roveri","sequence":"additional","affiliation":[{"name":"Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB), Politecnico di Milano, Milan, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"BranchyNet: Fast inference via early exiting from deep neural networks","author":"Teerapittayanon","year":"2017","journal-title":"arXiv:1709.01686"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1007\/s12559-020-09734-4"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1088\/1742-6596\/1168\/2\/022022"},{"key":"ref4","article-title":"IDK cascades: Fast deep learning by learning not to overthink","author":"Wang","year":"2017","journal-title":"arXiv:1706.00885"},{"key":"ref5","article-title":"Shallow-deep networks: Understanding and mitigating network overthinking","author":"Kaya","year":"2018","journal-title":"arXiv:1810.07052"},{"key":"ref6","article-title":"On the difficulty of training recurrent neural networks","author":"Pascanu","year":"2012","journal-title":"arXiv:1211.5063"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.l803.03635"},{"key":"ref8","first-page":"248","article-title":"ImageNet: A large-scale hierarchical image database","volume-title":"Proc. IEEE Conf. Comput. Vis. Pattern Recognit.","author":"Deng"},{"key":"ref9","first-page":"4167","article-title":"Differentiable branching in deep networks for fast inference","volume-title":"Proc. IEEE Int. Conf. Acoust., Speech Signal Process. (ICASSP)","author":"Scardapane"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-eacl.180"},{"key":"ref11","first-page":"500","article-title":"BADGE: Speeding up BERT inference after deployment via block-wise bypasses and divergence-based early exiting","volume-title":"Proc. 61st Annu. Meeting Assoc. Comput. Linguistics","author":"Zhu"},{"key":"ref12","doi-asserted-by":"crossref","DOI":"10.1016\/j.engappai.2023.106035","article-title":"Do we need early exit networks in human activity recognition?","volume":"121","author":"Lattanzi","year":"2023","journal-title":"Eng. Appl. Artif. Intell."},{"key":"ref13","article-title":"Temporal early exits for efficient video object detection","author":"Sabet","year":"2021","journal-title":"arXiv:2106.11208"},{"key":"ref14","article-title":"DynamicViT: Efficient vision transformers with dynamic token sparsification","author":"Rao","year":"2021","journal-title":"arXiv:2106.02034"},{"key":"ref15","first-page":"124","article-title":"Big\/little deep neural network for ultra low power inference","volume-title":"Proc. 10th Int. Conf. Hardw.\/Softw. Codesign Syst. Synth.","author":"Park"},{"key":"ref16","first-page":"1321","article-title":"On calibration of modern neural networks","volume-title":"Proc. 34th Intl. Conf. Mach. Learn.","author":"Guo"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1145\/3447582"},{"issue":"3","key":"ref18","doi-asserted-by":"crossref","first-page":"347","DOI":"10.1016\/0167-8191(90)90086-O","article-title":"Genetic algorithms and neural networks: Optimizing connections and connectivity","volume":"14","author":"Whitley","year":"1990","journal-title":"Parallel Comput."},{"key":"ref19","doi-asserted-by":"crossref","first-page":"97","DOI":"10.1016\/j.engappai.2017.01.013","article-title":"Metaheuristic design of feedforward neural networks: A review of two decades of research","volume":"60","author":"Ojha","year":"2017","journal-title":"Eng. Appl. Artif. Intell."},{"key":"ref20","first-page":"4322","article-title":"Hardware-aware neural architecture search: Survey and taxonomy","volume-title":"Proc. 30th Int. Joint Conf. Artif. Intell.","author":"Benmeziane"},{"key":"ref21","first-page":"2918","article-title":"CNAS: Constrained neural architecture search","volume-title":"Proc. IEEE Int. Conf. Syst., Man, Cybern. (SMC)","author":"Gambella"},{"key":"ref22","article-title":"MicroNAS: Memory and latency constrained hardware-aware neural architecture search for time series classification on microcontrollers","author":"King","year":"2023","journal-title":"arXiv:2310.18384"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.23919\/DATE56975.2023.10137095"},{"issue":"8","key":"ref24","doi-asserted-by":"crossref","first-page":"3832","DOI":"10.1109\/TNNLS.2021.3123105","article-title":"Automatic design of convolutional neural network architectures under resource constraints","volume":"34","author":"Li","year":"2023","journal-title":"IEEE Trans. Neural Netw. Learn. Syst."},{"key":"ref25","first-page":"1","article-title":"EExNAS: Early-exit neural architecture search solutions for low-power wearable devices","volume-title":"Proc. IEEE\/ACM Int. Symp. Low Power Electron. Design (ISLPED)","author":"Odema"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/TETCI.2024.3485677"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/IJCNN54540.2023.10191876"},{"key":"ref28","article-title":"Once-for-all: Train one network and specialize it for efficient deployment","author":"Cai","year":"2019","journal-title":"arXiv:1908.09791"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/IJCNN54540.2023.10191144"},{"issue":"2","key":"ref30","doi-asserted-by":"crossref","first-page":"182","DOI":"10.1109\/4235.996017","article-title":"A fast and elitist multiobjective genetic algorithm: NSGA-II","volume":"6","author":"Deb","year":"2002","journal-title":"IEEE Trans. Evol. Comput."},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1145\/3581791.3596845"},{"issue":"11","key":"ref32","doi-asserted-by":"crossref","first-page":"5475","DOI":"10.1109\/TNNLS.2018.2805098","article-title":"Deep cascade learning","volume":"29","author":"Marquez","year":"2018","journal-title":"IEEE Trans. Neural Netw. Learn. Syst."},{"key":"ref33","article-title":"Modeling teacher\u2013student techniques in deep neural networks for knowledge distillation","author":"Abbasi","year":"2019","journal-title":"arXiv:1912.13179"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3240765.3240796"},{"issue":"1","key":"ref35","doi-asserted-by":"crossref","first-page":"1","DOI":"10.3390\/e24010001","article-title":"A probabilistic re-intepretation of confidence scores in multi-exit models","volume":"24","author":"Pomponi","year":"2021","journal-title":"Entropy"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1905.01392"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00140"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1109\/TNN.1998.712192"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1613\/jair.301"},{"key":"ref40","article-title":"DARTS: Differentiable architecture search","author":"Liu","year":"2018","journal-title":"arXiv:1806.09055"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1887\/0750308958"},{"key":"ref42","volume-title":"Evolutionary Computation for Modeling and Optimization","author":"Ashlock","year":"2010"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1109\/tnnls.2021.3100554"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1109\/ICECCE49384.2020.9179174"},{"issue":"11","key":"ref45","doi-asserted-by":"crossref","first-page":"1245","DOI":"10.1016\/S0045-7825(01)00323-1","article-title":"Theoretical and numerical constraint-handling techniques used with evolutionary algorithms: A survey of the state of the art","volume":"191","author":"Coello Coello","year":"2002","journal-title":"Comput. Methods Appl. Mech. Eng."},{"key":"ref46","article-title":"NASViT: Neural architecture search for efficient vision transformers with gradient conflict aware supernet training","volume-title":"Proc. Int. Conf. Learn. Represent. (ICLR)","author":"Gong"},{"key":"ref47","article-title":"ChamNet: Towards efficient network design through platform-aware model adaptation","author":"Dai","year":"2018","journal-title":"arXiv:1812.08934"},{"key":"ref48","article-title":"Accelerating neural architecture search using performance prediction","author":"Baker","year":"2017","journal-title":"arXiv:1705.10823"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58452-8_3"},{"issue":"1","key":"ref50","doi-asserted-by":"crossref","first-page":"81","DOI":"10.1093\/biomet\/30.1-2.81","article-title":"A new measure of rank correlation","volume":"30","author":"Kendall","year":"1938","journal-title":"Biometrika"},{"key":"ref51","article-title":"Network morphism","author":"Wei","year":"2016","journal-title":"arXiv:1603.01670"},{"key":"ref52","doi-asserted-by":"crossref","first-page":"107","DOI":"10.1016\/j.ins.2020.02.041","article-title":"Optimized training and scalable implementation of conditional deep neural networks with early exits for fog-supported IoT applications","volume":"521","author":"Baccarelli","year":"2020","journal-title":"Inf. Sci."},{"key":"ref53","article-title":"Learning multiple layers of features from tiny images","author":"Krizhevsky","year":"2009"},{"key":"ref54","article-title":"CINIC-10 is not ImageNet or CIFAR-10","author":"Darlow","year":"2018","journal-title":"arXiv:1810.03505"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.2118\/18761-MS"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.3390\/INFOo11020108"},{"key":"ref57","article-title":"Best practices for scientific research on neural architecture search","author":"Lindauer","year":"2019","journal-title":"arXiv:1909.02453"},{"key":"ref58","first-page":"1","article-title":"Reducing the computation load of convolutional neural networks through gate classification","volume-title":"Proc. Int. Joint Conf. Neural Netw. (IJCNN)","author":"Disabato"},{"key":"ref59","article-title":"Deep residual learning for image recognition","author":"He","year":"2015","journal-title":"arXiv:1512.03385"},{"key":"ref60","article-title":"EfficientNet: Rethinking model scaling for convolutional neural networks","author":"Tan","year":"2019","journal-title":"arXiv:1905.11946"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v29i1.9602"},{"key":"ref62","article-title":"On calibration of modern neural networks","author":"Guo","year":"2017","journal-title":"arXiv:1706.04599"}],"container-title":["IEEE Transactions on Neural Networks and Learning Systems"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/5962385\/11195929\/11086509.pdf?arnumber=11086509","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,8]],"date-time":"2025-10-08T17:39:03Z","timestamp":1759945143000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11086509\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10]]},"references-count":62,"journal-issue":{"issue":"10"},"URL":"https:\/\/doi.org\/10.1109\/tnnls.2025.3588558","relation":{},"ISSN":["2162-237X","2162-2388"],"issn-type":[{"value":"2162-237X","type":"print"},{"value":"2162-2388","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10]]}}}