{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,12,31]],"date-time":"2025-12-31T00:40:09Z","timestamp":1767141609106,"version":"build-2238731810"},"reference-count":43,"publisher":"Springer Science and Business Media LLC","issue":"10","license":[{"start":{"date-parts":[[2024,9,24]],"date-time":"2024-09-24T00:00:00Z","timestamp":1727136000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2024,9,24]],"date-time":"2024-09-24T00:00:00Z","timestamp":1727136000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Sci. China Inf. Sci."],"published-print":{"date-parts":[[2024,10]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    Artificial intelligence (AI) has experienced substantial advancements recently, notably with the advent of large-scale language models (LLMs) employing mixture-of-experts (MoE) techniques, exhibiting human-like cognitive skills. As a promising hardware solution for edge MoE implementations, the computing-in-memory (CIM) architecture collocates memory and computing within a single device, significantly reducing the data movement and the associated energy consumption. However, due to diverse edge application scenarios and constraints, determining the optimal network structures for MoE, such as the expert\u2019s location, quantity, and dimension on CIM systems remains elusive. To this end, we introduce a software-hardware co-designed neural architecture search (NAS) framework,\n                    <jats:bold>C<\/jats:bold>\n                    IM-based\n                    <jats:bold>M<\/jats:bold>\n                    oE\n                    <jats:bold>N<\/jats:bold>\n                    AS (CMN), focusing on identifying a high-performing MoE structure under specific hardware constraints. The results of the NYUD-v2 dataset segmentation on the RRAM (SRAM) CIM system reveal that CMN can discover optimized MoE configurations under energy, latency, and performance constraints, achieving\n                    <jats:bold>29.67<\/jats:bold>\n                    \u00d7 (\n                    <jats:bold>43.10<\/jats:bold>\n                    \u00d7) energy savings,\n                    <jats:bold>175.44<\/jats:bold>\n                    \u00d7(\n                    <jats:bold>109.89<\/jats:bold>\n                    \u00d7) speedup, and\n                    <jats:bold>12.24<\/jats:bold>\n                    \u00d7 smaller model size compared to the baseline MoE-enabled Visual Transformer, respectively. This co-design opens up an avenue toward high-performance MoE deployments in edge CIM systems.\n                  <\/jats:p>","DOI":"10.1007\/s11432-024-4144-y","type":"journal-article","created":{"date-parts":[[2024,9,26]],"date-time":"2024-09-26T05:02:13Z","timestamp":1727326933000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":2,"title":["CMN: a co-designed neural architecture search for efficient computing-in-memory-based mixture-of-experts"],"prefix":"10.1007","volume":"67","author":[{"given":"Shihao","family":"Han","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sishuo","family":"Liu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Shucheng","family":"Du","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mingzi","family":"Li","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zijian","family":"Ye","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xiaoxin","family":"Xu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yi","family":"Li","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Zhongrui","family":"Wang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Dashan","family":"Shang","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","published-online":{"date-parts":[[2024,9,24]]},"reference":[{"key":"4144_CR1","volume-title":"Improving language understanding by generative pre-training","author":"A Radford","year":"2018","unstructured":"Radford A, Narasimhan K, Salimans T, et al. Improving language understanding by generative pre-training. 2018. https:\/\/cdn.openai.com\/research-covers\/language-unsupervised\/language_understanding_paper.pdf"},{"key":"4144_CR2","unstructured":"Brown T, Mann B, Ryder N, et al. Language models are few-shot learners. arXiv:2005.14165, 2020"},{"key":"4144_CR3","unstructured":"Devlin J, Chang M W, Lee K, et al. BERT: pre-training of deep bidirectional transformers for language understanding. arXiv:1810.04805, 2018"},{"key":"4144_CR4","unstructured":"Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need. arXiv:1706.03762, 2017"},{"key":"4144_CR5","first-page":"1","volume":"23","author":"W Fedus","year":"2022","unstructured":"Fedus W, Zoph B, Shazeer N. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. J Mach Learn Res, 2022, 23: 1\u201339","journal-title":"J Mach Learn Res"},{"key":"4144_CR6","unstructured":"Shazeer N, Mirhoseini A, Maziarz K, et al. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. arXiv:1701.06538, 2017"},{"key":"4144_CR7","unstructured":"Zhang S, Roller S, Goyal N, et al. Opt: Open pre-trained transformer language models. arXiv:2205.01068, 2022"},{"key":"4144_CR8","doi-asserted-by":"publisher","first-page":"101","DOI":"10.1038\/nmat4756","volume":"16","author":"Z R Wang","year":"2016","unstructured":"Wang Z R, Joshi S, Savel\u2019ev S, et al. Memristors with diffusive dynamics as synaptic emulators for neuromorphic computing. Nat Mater, 2016, 16: 101\u2013108","journal-title":"Nat Mater"},{"key":"4144_CR9","doi-asserted-by":"publisher","first-page":"2632","DOI":"10.1002\/adma.200900375","volume":"21","author":"R Waser","year":"2009","unstructured":"Waser R, Dittmann R, Staikov G, et al. Redox-based resistive switching memories-nanoionic mechanisms, prospects, and challenges. Adv Mater, 2009, 21: 2632\u20132663","journal-title":"Adv Mater"},{"key":"4144_CR10","doi-asserted-by":"publisher","first-page":"063002","DOI":"10.1088\/0268-1242\/31\/6\/063002","volume":"31","author":"D Ielmini","year":"2016","unstructured":"Ielmini D. Resistive switching memories based on metal oxides: mechanisms, reliability and scaling. Semicon Sci Technol, 2016, 31: 063002","journal-title":"Semicon Sci Technol"},{"key":"4144_CR11","doi-asserted-by":"publisher","first-page":"333","DOI":"10.1038\/s41928-018-0092-2","volume":"1","author":"D Ielmini","year":"2018","unstructured":"Ielmini D, Wong H S P. In-memory computing with resistive switching devices. Nat Electron, 2018, 1: 333\u2013343","journal-title":"Nat Electron"},{"key":"4144_CR12","doi-asserted-by":"publisher","first-page":"841","DOI":"10.1038\/nmat2028","volume":"6","author":"W Lu","year":"2007","unstructured":"Lu W, Lieber C M. Nanoelectronics from the bottom up. Nat Mater, 2007, 6: 841\u2013850","journal-title":"Nat Mater"},{"key":"4144_CR13","first-page":"1871","volume":"70","author":"Y Li","year":"2023","unstructured":"Li Y, Chen J, Wang L F, et al. An ADC-less rram-based computing-in-memory macro with binary CNN for efficient edge AI. IEEE Trans Circu Syst II: Expr Brief, 2023, 70: 1871\u20131875","journal-title":"IEEE Trans Circu Syst II: Expr Brief"},{"key":"4144_CR14","doi-asserted-by":"publisher","first-page":"264","DOI":"10.1016\/j.eng.2020.01.007","volume":"6","author":"Y R Chen","year":"2020","unstructured":"Chen Y R, Xie Y, Song L H, et al. A survey of accelerator architectures for deep neural networks. Engineering, 2020, 6: 264\u2013274","journal-title":"Engineering"},{"key":"4144_CR15","doi-asserted-by":"publisher","first-page":"100","DOI":"10.3390\/fi11040100","volume":"11","author":"M Capra","year":"2019","unstructured":"Capra M, Peloso R, Masera G, et al. Edge computing: A survey on the hardware requirements in the internet of things world. Future Internet, 2019, 11: 100","journal-title":"Future Internet"},{"key":"4144_CR16","doi-asserted-by":"publisher","first-page":"420","DOI":"10.1038\/s41928-019-0288-0","volume":"2","author":"W H Chen","year":"2019","unstructured":"Chen W H, Dou C M, Li K X, et al. Cmos-integrated memristive non-volatile computing-in-memory for AI edge processors. Nature Electron, 2019, 2: 420\u2013428","journal-title":"Nature Electron"},{"key":"4144_CR17","doi-asserted-by":"publisher","first-page":"2473","DOI":"10.1038\/s41467-020-16108-9","volume":"11","author":"V Joshi","year":"2020","unstructured":"Joshi V, Le Gallo M, Haefeli S, et al. Accurate deep neural network inference using computational phasechange memory. Nature Commun, 2020, 11: 2473","journal-title":"Nature Commun"},{"key":"4144_CR18","doi-asserted-by":"publisher","first-page":"2468","DOI":"10.1038\/s41467-021-22364-0","volume":"12","author":"G Karunaratne","year":"2021","unstructured":"Karunaratne G, Schmuck M, Le Gallo M, et al. Robust high-dimensional memory-augmented neural networks. Nature Commun, 2021, 12: 2468","journal-title":"Nature Commun"},{"key":"4144_CR19","doi-asserted-by":"publisher","first-page":"480","DOI":"10.1038\/s41928-019-0313-3","volume":"2","author":"J Moon","year":"2019","unstructured":"Moon J, Ma W, Shin J H, et al. Temporal data classification and forecasting using a memristor-based reservoir computing system. Nature Electron, 2019, 2: 480\u2013487","journal-title":"Nature Electron"},{"key":"4144_CR20","doi-asserted-by":"publisher","first-page":"529","DOI":"10.1038\/s41565-020-0655-z","volume":"15","author":"A Sebastian","year":"2020","unstructured":"Sebastian A, Le Gallo M, Khaddam-Aljameh R, et al. Memory devices and applications for in-memory computing. Nature Nanotech, 2020, 15: 529\u2013544","journal-title":"Nature Nanotech"},{"key":"4144_CR21","doi-asserted-by":"publisher","first-page":"80","DOI":"10.1038\/nature06932","volume":"453","author":"D Strukov","year":"2008","unstructured":"Strukov D, Snider G, Stewart D, et al. The missing memristor found. Nature, 2008, 453: 80\u201383","journal-title":"Nature"},{"key":"4144_CR22","doi-asserted-by":"publisher","first-page":"371","DOI":"10.1038\/s41928-020-0435-7","volume":"3","author":"W Q Zhang","year":"2020","unstructured":"Zhang W Q, Gao B, Tang J S, et al. Neuro-inspired computing chips. Nature Electron, 2020, 3: 371\u2013382","journal-title":"Nature Electron"},{"key":"4144_CR23","doi-asserted-by":"publisher","first-page":"2200026","DOI":"10.1002\/aisy.202200026","volume":"4","author":"Y Li","year":"2022","unstructured":"Li Y, Zhang W Y, Xu X X, et al. Mixed-precision continual learning based on computational resistance random access memory. Adv Intell Syst, 2022, 4: 2200026","journal-title":"Adv Intell Syst"},{"key":"4144_CR24","doi-asserted-by":"crossref","unstructured":"Jawahar G, Mukherjee S, Liu X D, et al. Automoe: Heterogeneous mixture-of-experts with adaptive computation for efficient neural machine translation. arXiv:2210.07535, 2022","DOI":"10.18653\/v1\/2023.findings-acl.580"},{"key":"4144_CR25","doi-asserted-by":"crossref","unstructured":"Wang H R, Wu Z H, Liu Z J, et al. Hat: Hardware-aware transformers for efficient natural language processing. arXiv:2005.14187, 2020","DOI":"10.18653\/v1\/2020.acl-main.686"},{"key":"4144_CR26","doi-asserted-by":"publisher","first-page":"1705914","DOI":"10.1002\/adma.201705914","volume":"30","author":"M Hu","year":"2018","unstructured":"Hu M, Graves C, Li C, et al. Memristor-based analog computation and neural network classification with a dot product engine. Adv Mater, 2018, 30: 1705914","journal-title":"Adv Mater"},{"key":"4144_CR27","doi-asserted-by":"publisher","first-page":"371","DOI":"10.1038\/s41928-020-0435-7","volume":"3","author":"W Q Zhang","year":"2020","unstructured":"Zhang W Q, Gao B, Tang J S, et al. Neuro-inspired computing chips. Nature Electron, 2020, 3: 371\u2013382","journal-title":"Nature Electron"},{"key":"4144_CR28","first-page":"252","volume":"64","author":"Y D Chih","year":"2021","unstructured":"Chih Y D, Lee P H, Fujiwara H, et al. 16.4 An 89 TOPS\/W and 16.3 TOPS\/mm2 all-digital SRAM-based full-precision compute-in memory macro in 22nm for machine-learning edge applications. In: IEEE International Solid-State Circuits Conference (ISSCC), 2021, 64: 252\u2013254","journal-title":"IEEE International Solid-State Circuits Conference (ISSCC)"},{"key":"4144_CR29","unstructured":"Zhuang J T. Laddernet: Multi-path networks based on u-net for medical image segmentation. arXiv:1810.07810, 2018"},{"key":"4144_CR30","first-page":"12077","volume":"34","author":"E Z Xie","year":"2021","unstructured":"Xie E Z, Wang W H, Yu Z D, et al. Segformer: Simple and efficient design for semantic segmentation with transformers. Adv Neural Inform Process Syst, 2021: 34: 12077\u201312090","journal-title":"Adv Neural Inform Process Syst"},{"key":"4144_CR31","unstructured":"Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv:2010.11929, 2020"},{"key":"4144_CR32","first-page":"760","volume-title":"Particle Swarm Optimization","author":"J Kennedy","year":"2010","unstructured":"Kennedy J. Particle Swarm Optimization. Springer US, Boston, MA, 2010. 760\u2013766"},{"key":"4144_CR33","doi-asserted-by":"publisher","first-page":"79","DOI":"10.1007\/s10462-016-9517-3","volume":"49","author":"A Nabaei","year":"2018","unstructured":"Nabaei A, Hamian M, Parsaei M, et al. Topologies and performance of intelligent algorithms: a comprehensive review. Artificial Intell Review, 2018, 49: 79\u2013103","journal-title":"Artificial Intell Review"},{"key":"4144_CR34","doi-asserted-by":"publisher","first-page":"641","DOI":"10.1038\/s41586-020-1942-4","volume":"577","author":"P Yao","year":"2020","unstructured":"Yao P, Wu H Q, Gao B, et al. Fully hardware-implemented memristor convolutional neural network. Nature, 2020, 577: 641\u2013646","journal-title":"Nature"},{"key":"4144_CR35","doi-asserted-by":"publisher","first-page":"325","DOI":"10.1126\/science.adf5538","volume":"384","author":"T H Wen","year":"2024","unstructured":"Wen T H, Hung J M, Huang W H, et al. Fusion of memristor and digital compute-in-memory processing for energy-efficient edge computing. Science, 2024, 384: 325\u2013332","journal-title":"Science"},{"key":"4144_CR36","doi-asserted-by":"publisher","first-page":"1902761","DOI":"10.1002\/adma.201902761","volume":"31","author":"J S Tang","year":"2019","unstructured":"Tang J S, Yuan F, Shen X K, et al. Bridging biological and artificial neural networks with emerging neuromorphic devices: Fundamentals, progress, and challenges. Advanced Mater, 2019, 31: 1902761","journal-title":"Advanced Mater"},{"key":"4144_CR37","unstructured":"Analog Devices. 1 MSPS Ultralow Power 12-Bit ADC in 10-Lead LFCSP and MSOP: AD7091R. 2016. Rev. B. https:\/\/www.analog.com\/media\/en\/technical-documentation\/data-sheets\/AD7091R.pdf"},{"key":"4144_CR38","first-page":"471","volume-title":"Proceedings of the ESSCIRC (ESSCIRC)","author":"R Sekimoto","year":"2011","unstructured":"Sekimoto R, Shikata A, Kuroda T, et al. A 40nm 50s\/s-8ms\/s ultra low voltage sar adc with timing optimized asynchronous clock generator. In: Proceedings of the ESSCIRC (ESSCIRC), 2011. 471\u2013474"},{"key":"4144_CR39","first-page":"236","volume-title":"IEEE International Solid-State Circuits Conference (ISSCC)","author":"P Vogelmann","year":"2018","unstructured":"Vogelmann P, Haas M, Ortmanns M. A 1.1mw 200kS\/s incremental \u0394\u03a3 ADC with a DR of 91.5dB using integrator slicing for dynamic power reduction. In: IEEE International Solid-State Circuits Conference (ISSCC), 2018. 236\u2013238"},{"key":"4144_CR40","doi-asserted-by":"publisher","first-page":"1369","DOI":"10.1364\/PRJ.420887","volume":"9","author":"AS Pal","year":"2021","unstructured":"Pal AS, Zhang Y, Yau D. Monolithic and single-functional-unit level integration of electronic and photonic elements: FET-LET hybrid 6T SRAM. Photonics Research, 2021, 9: 1369\u20131378","journal-title":"Photonics Research"},{"key":"4144_CR41","unstructured":"Houshmand P, Sun J C, Verhelst M. Benchmarking and modeling of analog and digital SRAM in-memory computing architectures. arXiv:2305.18335, 2023"},{"key":"4144_CR42","doi-asserted-by":"publisher","first-page":"2204944","DOI":"10.1002\/adma.202204944","volume":"35","author":"W Haensch","year":"2023","unstructured":"Haensch W, Raghunathan A, Roy K, et al. Compute in-memory with non-volatile elements for neural networks: A review from a co-design perspective. Advanced Mater, 2023, 35: 2204944","journal-title":"Advanced Mater"},{"key":"4144_CR43","first-page":"1","volume-title":"IEEE 3rd International Conference on Artificial Intelligence Circuits and Systems (AICAS)","author":"A Lu","year":"2021","unstructured":"Lu A, Peng X C, LiW T, et al. Neurosim validation with 40nm RRAM compute-in-memory macro. In: IEEE 3rd International Conference on Artificial Intelligence Circuits and Systems (AICAS), 2021. 1\u20134"}],"updated-by":[{"DOI":"10.1007\/s11432-024-4168-5","type":"erratum","label":"Erratum","source":"publisher","updated":{"date-parts":[[2024,10,18]],"date-time":"2024-10-18T00:00:00Z","timestamp":1729209600000}}],"container-title":["Science China Information Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11432-024-4144-y.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s11432-024-4144-y\/fulltext.html","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11432-024-4144-y.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,19]],"date-time":"2025-11-19T20:24:18Z","timestamp":1763583858000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s11432-024-4144-y"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,9,24]]},"references-count":43,"journal-issue":{"issue":"10","published-print":{"date-parts":[[2024,10]]}},"alternative-id":["4144"],"URL":"https:\/\/doi.org\/10.1007\/s11432-024-4144-y","relation":{},"ISSN":["1674-733X","1869-1919"],"issn-type":[{"value":"1674-733X","type":"print"},{"value":"1869-1919","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,9,24]]},"assertion":[{"value":"5 April 2024","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"29 July 2024","order":2,"name":"revised","label":"Revised","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"2 September 2024","order":3,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"24 September 2024","order":4,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"21 October 2024","order":6,"name":"change_date","label":"Change Date","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"Erratum","order":7,"name":"change_type","label":"Change Type","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"An Erratum to this paper has been published:","order":8,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"https:\/\/doi.org\/10.1007\/s11432-024-4168-5","URL":"https:\/\/doi.org\/10.1007\/s11432-024-4168-5","order":9,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"18 October 2024","order":10,"name":"change_date","label":"Change Date","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"Erratum","order":11,"name":"change_type","label":"Change Type","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"An Erratum to this paper has been published:","order":12,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"https:\/\/doi.org\/10.1007\/s11432-024-4168-5","URL":"https:\/\/doi.org\/10.1007\/s11432-024-4168-5","order":13,"name":"change_details","label":"Change Details","group":{"name":"ArticleHistory","label":"Article History"}}],"article-number":"200405"}}