{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,1]],"date-time":"2026-08-01T16:51:54Z","timestamp":1785603114373,"version":"3.56.0"},"reference-count":117,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"12","license":[{"start":{"date-parts":[[2024,12,1]],"date-time":"2024-12-01T00:00:00Z","timestamp":1733011200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2024,12,1]],"date-time":"2024-12-01T00:00:00Z","timestamp":1733011200000},"content-version":"am","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2024,12,1]],"date-time":"2024-12-01T00:00:00Z","timestamp":1733011200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2024,12,1]],"date-time":"2024-12-01T00:00:00Z","timestamp":1733011200000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"name":"NSF","award":["CNS 2007284"],"award-info":[{"award-number":["CNS 2007284"]}]},{"name":"iMAGiNE"},{"name":"NSF","award":["2133861"],"award-info":[{"award-number":["2133861"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Pattern Anal. Mach. Intell."],"published-print":{"date-parts":[[2024,12]]},"DOI":"10.1109\/tpami.2024.3395423","type":"journal-article","created":{"date-parts":[[2024,4,30]],"date-time":"2024-04-30T19:13:53Z","timestamp":1714504433000},"page":"7618-7635","source":"Crossref","is-referenced-by-count":46,"title":["Zero-Shot Neural Architecture Search: Challenges, Solutions, and Opportunities"],"prefix":"10.1109","volume":"46","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-8537-8632","authenticated-orcid":false,"given":"Guihong","family":"Li","sequence":"first","affiliation":[{"name":"Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4512-8465","authenticated-orcid":false,"given":"Duc","family":"Hoang","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8115-4276","authenticated-orcid":false,"given":"Kartikeya","family":"Bhardwaj","sequence":"additional","affiliation":[{"name":"Qualcomm AI Research, Qualcomm Technologies, Inc., San Diego, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5741-0516","authenticated-orcid":false,"given":"Ming","family":"Lin","sequence":"additional","affiliation":[{"name":"Amazon, Seattle, WA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2050-5693","authenticated-orcid":false,"given":"Zhangyang","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1826-7646","authenticated-orcid":false,"given":"Radu","family":"Marculescu","sequence":"additional","affiliation":[{"name":"Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, TX, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","first-page":"1106","article-title":"ImageNet classification with deep convolutional neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Krizhevsky"},{"key":"ref2","doi-asserted-by":"publisher","DOI":"10.1109\/ACPR.2015.7486599"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.243"},{"key":"ref4","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"ref5","article-title":"An image is worth 16x16 words: Transformers for image recognition at scale","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Dosovitskiy"},{"key":"ref6","article-title":"Language models are few-shot learners","author":"Brown","year":"2020","journal-title":"arXiv: 2005.14165"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1706.03762"},{"key":"ref8","article-title":"Designing neural network architectures using reinforcement learning","author":"Baker","year":"2016","journal-title":"arXiv:1611.02167"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1611.01578"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01246-5_2"},{"key":"ref11","article-title":"DARTS: Differentiable architecture search","author":"Liu","year":"2018","journal-title":"arXiv: 1806.09055"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-05318-5_3"},{"key":"ref13","first-page":"4095","article-title":"Efficient neural architecture search via parameters sharing","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Pham"},{"key":"ref14","first-page":"2902","article-title":"Large-scale evolution of image classifiers","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Real"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00332"},{"key":"ref16","article-title":"SNAS: Stochastic neural architecture search","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Xie"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.01099"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.01298"},{"key":"ref19","first-page":"367","article-title":"Random search and reproducibility for neural architecture search","volume-title":"Proc. 35th Conf. Uncertainty Artif. Intell.","author":"Li"},{"key":"ref20","first-page":"2020","article-title":"Neural architecture search with Bayesian optimisation and optimal transport","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Kandasamy"},{"key":"ref21","article-title":"Evaluating the search phase of neural architecture search","author":"Yu","year":"2019","journal-title":"arXiv: 1902.08142"},{"key":"ref22","article-title":"Hierarchical representations for efficient architecture search","author":"Liu","year":"2017","journal-title":"arXiv: 1711.00436"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v32i1.11709"},{"key":"ref24","first-page":"7827","article-title":"Neural architecture optimization","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Luo"},{"key":"ref25","article-title":"Graph hypernetworks for neural architecture search","author":"Zhang","year":"2018","journal-title":"arXiv: 1810.05749"},{"key":"ref26","first-page":"7603","article-title":"BayesNAS: A Bayesian approach for neural architecture search","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zhou"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00140"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58571-6_41"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00293"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.1145\/3341302.3342080"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1145\/3292500.3330925"},{"key":"ref32","article-title":"PC-DARTS: Partial channel connections for memory-efficient architecture search","author":"Xu","year":"2019","journal-title":"arXiv: 1907.05737"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00186"},{"key":"ref34","article-title":"Understanding and robustifying differentiable architecture search","author":"Zela","year":"2019","journal-title":"arXiv: 1909.09656"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00138"},{"key":"ref36","article-title":"ProxylessNAS: Direct neural architecture search on target task and hardware","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Cai"},{"key":"ref37","article-title":"Once-for-all: Train one network and specialize it for efficient deployment","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Cai"},{"key":"ref38","article-title":"Neural architecture search on ImageNet","year":"2023"},{"key":"ref39","article-title":"Single-path NAS: Designing hardware-efficient ConvNets in less than 4 hours","author":"Stamoulis","year":"2019","journal-title":"arXiv: 1904.02877"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.01202"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58517-4_32"},{"key":"ref42","article-title":"FasterSeg: Searching for faster real-time semantic segmentation","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Chen"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.1145\/3491396.3506510"},{"key":"ref44","article-title":"Unifying and boosting gradient-based training-free neural architecture search","author":"Shu","year":"2022","journal-title":"arXiv:2201.09785"},{"key":"ref45","article-title":"LiteTransformerSearch: Training-free on-device search for efficient autoregressive language models","author":"Javaheripi","year":"2022","journal-title":"arXiv:2203.02094"},{"key":"ref46","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01062"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1145\/3528416.3530873"},{"issue":"7","key":"ref48","first-page":"855","article-title":"Training-free hardware-aware neural architecture search with reinforcement learning","volume":"26","author":"Tran","year":"2021","journal-title":"J. Broadcast Eng."},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2021.3115911"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-92270-2_29"},{"key":"ref51","article-title":"Zero-cost proxies meet differentiable architecture search","author":"Xiang","year":"2021","journal-title":"arXiv:2106.06799"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.01141"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01329"},{"key":"ref54","article-title":"Neural architecture search on imagenet in four GPU hours: A theoretically inspired perspective","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Chen"},{"key":"ref55","article-title":"Zero-cost proxies for lightweight NAS","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Abdelfattah"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.1016\/j.knosys.2020.106622"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1905.01392"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1145\/3447582"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2021.3100554"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.1145\/3473330"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.24963\/ijcai.2021\/592"},{"key":"ref62","article-title":"A deeper look at zero-cost proxies for lightweight nas","volume-title":"Proc. ICLR Blog Track","author":"White"},{"key":"ref63","first-page":"12265","article-title":"Evaluating efficient performance estimators of neural architectures","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Ning"},{"key":"ref64","first-page":"28454","article-title":"How powerful are performance predictors in neural architecture search?","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"White"},{"key":"ref65","article-title":"Understanding and accelerating neural architecture search with training-free and theory-grounded metrics","author":"Chen","year":"2021","journal-title":"arXiv:2108.11939"},{"key":"ref66","first-page":"14\/1","article-title":"\u201cNo free lunch","volume-title":"Proc. Automated Mach. Learn. Conf.","author":"Chen"},{"key":"ref67","article-title":"On the expressive power of overlapping architectures of deep learning","volume-title":"Proc. 6th Int. Conf. Learn. Representations","author":"Sharir"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.1007\/s10115-021-01605-0"},{"key":"ref69","article-title":"Generalization in deep networks: The role of distance from initialization","author":"Nagarajan","year":"2019","journal-title":"arXiv: 1901.01672"},{"key":"ref70","article-title":"Towards understanding generalization of deep learning: Perspective of loss landscapes","author":"Wu","year":"2017","journal-title":"arXiv: 1706.10239"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1109\/TNNLS.2018.2868980"},{"key":"ref72","first-page":"10462","article-title":"Disentangling trainability and generalization in deep neural networks","volume-title":"Proc. 37th Int. Conf. Mach. Learn.","author":"Xiao"},{"key":"ref73","first-page":"11611","article-title":"Uniform convergence may be unable to explain generalization in deep learning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Nagarajan"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2022.06.031"},{"key":"ref75","article-title":"SNIP: Single-shot network pruning based on connection sensitivity","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Lee"},{"key":"ref76","first-page":"6377","article-title":"Pruning neural networks without any data by iteratively conserving synaptic flow","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Tanaka"},{"key":"ref77","article-title":"Picking winning tickets before training by preserving gradient flow","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Wang"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.01152"},{"key":"ref79","article-title":"GradSign: Model performance inference with theoretical insights","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Zhang"},{"key":"ref80","article-title":"Faster gaze prediction with dense networks and fisher pruning","author":"Theis","year":"2018","journal-title":"arXiv: 1801.05787"},{"key":"ref81","first-page":"7021","article-title":"Group fisher pruning for practical network compression","volume-title":"Proc. 38th Int. Conf. Mach. Learn.","author":"Liu"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-86383-8_44"},{"key":"ref83","first-page":"7588","article-title":"Neural architecture search without training","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Mellor"},{"key":"ref84","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00040"},{"key":"ref85","first-page":"20810","article-title":"MAE- DET: Revisiting maximum entropy principle in zero-shot NAS for efficient object detection","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Sun"},{"key":"ref86","first-page":"2933","article-title":"On lazy training in differentiable programming","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Chizat"},{"key":"ref87","first-page":"8580","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Jacot"},{"key":"ref88","first-page":"8570","article-title":"Wide neural networks of any depth evolve as linear models under gradient descent","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Lee"},{"key":"ref89","article-title":"Auto-scaling vision transformers without training","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Chen"},{"key":"ref90","first-page":"2847","article-title":"On the expressive power of deep neural networks","volume-title":"Proc. 34th Int. Conf. Mach. Learn. Sydney","author":"Raghu"},{"key":"ref91","first-page":"4565","article-title":"Bounding and counting linear regions of deep neural networks","volume-title":"Proc. 35th Int. Conf. Mach. Learn.","author":"Serra"},{"key":"ref92","first-page":"10514","article-title":"On the number of linear regions of convolutional neural networks","volume-title":"Proc. 37th Int. Conf. Mach. Learn.","author":"Xiong"},{"key":"ref93","first-page":"2596","article-title":"Complexity of linear regions in deep networks","volume-title":"Proc. 36th Int. Conf. Mach. Learn.","author":"Hanin"},{"key":"ref94","article-title":"Restructurable activation networks","author":"Bhardwaj","year":"2022","journal-title":"arXiv:2208.08562"},{"issue":"5s","key":"ref95","doi-asserted-by":"crossref","first-page":"63:1","DOI":"10.1145\/3476994","article-title":"FLASH: Fast neural architecture search with hardware optimization","volume":"20","author":"Li","year":"2021","journal-title":"ACM Trans. Embedded Comput. Syst."},{"key":"ref96","first-page":"35298","article-title":"Deep architecture connectivity matters for its convergence: A fine-grained analysis","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Chen"},{"key":"ref97","article-title":"Deep neural networks as Gaussian processes","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Lee"},{"key":"ref98","article-title":"NAS-bench-301 and the case for surrogate benchmarks for neural architecture search","author":"Siems","year":"2020","journal-title":"arXiv: 2008.09777"},{"key":"ref99","article-title":"NAS-bench-suite: NAS evaluation is (now) surprisingly easy","author":"Mehta","year":"2022","journal-title":"arXiv:2201.13396"},{"key":"ref100","article-title":"NAS-bench-ASR: Reproducible neural architecture search for speech recognition","volume-title":"Proc. Int. Conf. Learn. Representations","author":"Mehrotra"},{"key":"ref101","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2022.3169897"},{"key":"ref102","first-page":"7105","article-title":"NAS-bench-101: Towards reproducible neural architecture search","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Ying"},{"key":"ref103","article-title":"NAS-Bench-201: Extending the scope of reproducible neural architecture search","author":"Dong","year":"2020","journal-title":"arXiv: 2001.00326"},{"key":"ref104","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2021.3054824"},{"key":"ref105","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00521"},{"key":"ref106","article-title":"HW-NAS-bench: Hardware-aware neural architecture search benchmark","author":"Li","year":"2021","journal-title":"arXiv:2103.10584"},{"key":"ref107","first-page":"10480","article-title":"BRP- NAS: Prediction-based NAS using GCNs","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Dudziak"},{"key":"ref108","first-page":"27016","article-title":"Hardware-adaptive efficient latency prediction for NAS via meta-learning","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Lee"},{"key":"ref109","doi-asserted-by":"publisher","DOI":"10.1145\/3458864.3467882"},{"key":"ref110","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2003.819861"},{"key":"ref111","article-title":"PyTorch image models","author":"Wightman","year":"2019"},{"key":"ref112","article-title":"NanoDet-plus: Super fast and high accuracy lightweight anchor-free object detection model","year":"2021"},{"key":"ref113","first-page":"6231","article-title":"The expressive power of neural networks: A view from the width","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Lu"},{"key":"ref114","doi-asserted-by":"publisher","DOI":"10.1016\/0893-6080(89)90020-8"},{"key":"ref115","first-page":"13883","article-title":"Overparameterization improves robustness to covariate shift in high dimensions","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","author":"Tripuraneni"},{"key":"ref116","first-page":"6155","article-title":"Learning and generalization in overparameterized neural networks, going beyond two layers","volume-title":"Proc. Adv. Neural Inf. Process. Syst.: Annu. Conf. Neural Inf. Process. Syst.","author":"Allen-Zhu"},{"key":"ref117","article-title":"ZiCo: Zero-shot NAS via inverse coefficient of variation on gradients","author":"Li","year":"2023","journal-title":"arXiv:2301.11300"}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/ieeexplore.ieee.org\/ielam\/34\/10746266\/10516268-aam.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"http:\/\/xplorestaging.ieee.org\/ielx7\/34\/10746266\/10516268.pdf?arnumber=10516268","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,11,27]],"date-time":"2024-11-27T00:10:43Z","timestamp":1732666243000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/10516268\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,12]]},"references-count":117,"journal-issue":{"issue":"12"},"URL":"https:\/\/doi.org\/10.1109\/tpami.2024.3395423","relation":{},"ISSN":["0162-8828","2160-9292","1939-3539"],"issn-type":[{"value":"0162-8828","type":"print"},{"value":"2160-9292","type":"electronic"},{"value":"1939-3539","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,12]]}}}