{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,17]],"date-time":"2026-06-17T16:25:59Z","timestamp":1781713559408,"version":"3.54.5"},"reference-count":34,"publisher":"Association for Computing Machinery (ACM)","issue":"3","license":[{"start":{"date-parts":[[2026,5,13]],"date-time":"2026-05-13T00:00:00Z","timestamp":1778630400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Embed. Comput. Syst."],"published-print":{"date-parts":[[2026,5,31]]},"abstract":"<jats:p>\n                    Neural network-based wireless receivers, also known as\n                    <jats:italic toggle=\"yes\">neural receivers<\/jats:italic>\n                    , have demonstrated superior performance over traditional receivers, but come with greater computational complexity. The need to use these networks on energy-conscious edge devices is increasing, necessitating energy-efficient and adaptive neural receivers to function well under varying channel conditions. Transitioning static neural receivers to dynamic models using the concepts of Dynamic Neural Networks (DyNN) could reduce runtime computational complexity and energy consumption. This could be achieved by adapting the network architecture during run-time and exploit the varying channel conditions in a mobile communication system to reduce the computational complexity. This work introduces MEAN, a novel hard-gated Mixture-of-Experts (MOE) based neural receiver architecture. The main idea behind MEAN is to use several smaller Signal-to-Noise-Ratio (SNR) expert networks are used during run-time to create a network that selects the correct expert for the current data input to reduce complexity. The paper consists of the following key contributions.\n                    <jats:list list-type=\"ordered\">\n                      <jats:list-item>\n                        <jats:label>(1)<\/jats:label>\n                        <jats:p>\n                          MEAN architecture based on\u00a0[\n                          <jats:xref ref-type=\"bibr\">25<\/jats:xref>\n                          ]: A hard-gated MoE model that dynamically selects the most suitable expert for each input dynamically based on current channel conditions.\n                        <\/jats:p>\n                      <\/jats:list-item>\n                      <jats:list-item>\n                        <jats:label>(2)<\/jats:label>\n                        <jats:p>Comprehensive system-level performance analysis of MEAN across different code rates and modulation schemes, including validation of expert selection during inference.<\/jats:p>\n                      <\/jats:list-item>\n                      <jats:list-item>\n                        <jats:label>(3)<\/jats:label>\n                        <jats:p>Loss function optimization for the gating network to promote the activation of specific experts in designated noise regions. This enhancement reduces gating network complexity while improving expert selection accuracy over previous MEAN implementations.<\/jats:p>\n                      <\/jats:list-item>\n                      <jats:list-item>\n                        <jats:label>(4)<\/jats:label>\n                        <jats:p>Hardware analysis of a gate-level implementation of MEAN synthesized in 22nm FD-SOI technology. The design is generated using a custom High-Level Synthesis (HLS) framework, providing detailed power and area evaluations.<\/jats:p>\n                      <\/jats:list-item>\n                    <\/jats:list>\n                  <\/jats:p>\n                  <jats:p>The proposed MEAN architecture for a Single Input Multiple Output (SIMO) wireless system achieves a 37.11% reduction in total power consumption with only an 11.04% increase in area, while maintaining accuracy comparable to that of a static neural network.<\/jats:p>","DOI":"10.1145\/3765519","type":"journal-article","created":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T11:24:33Z","timestamp":1756725873000},"page":"1-23","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":3,"title":["MEAN: Mixture-of-Experts Neural Receiver - Architecture and Performance Analysis"],"prefix":"10.1145","volume":"25","author":[{"ORCID":"https:\/\/orcid.org\/0009-0002-5393-2031","authenticated-orcid":false,"given":"Bram","family":"van Bolderik","sequence":"first","affiliation":[{"name":"Electronic systems, Eindhoven University of Technology","place":["Eindhoven, Netherlands"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5262-0605","authenticated-orcid":false,"given":"Vlado","family":"Menkovski","sequence":"additional","affiliation":[{"name":"Electronic systems, Eindhoven University of Technology","place":["Eindhoven, Netherlands"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2270-727X","authenticated-orcid":false,"given":"Sonia","family":"Heemstra de Groot","sequence":"additional","affiliation":[{"name":"Electronic systems, Eindhoven University of Technology","place":["Eindhoven, Netherlands"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5889-0785","authenticated-orcid":false,"given":"Manil Dev","family":"Gomony","sequence":"additional","affiliation":[{"name":"Electronic systems, Eindhoven University of Technology","place":["Eindhoven, Netherlands"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2026,5,13]]},"reference":[{"key":"e_1_3_1_2_2","doi-asserted-by":"publisher","DOI":"10.1109\/ASP-DAC58780.2024.10473894"},{"key":"e_1_3_1_3_2","doi-asserted-by":"publisher","DOI":"10.1109\/TWC.2021.3101364"},{"key":"e_1_3_1_4_2","unstructured":"Cadence. 2022. HLS for AI\/ML models: TensorFlow to RTL. (2022)."},{"key":"e_1_3_1_5_2","unstructured":"Shaofeng Cai Yao Shu Wei Wang and Beng Chin Ooi. 2019. ISBNet: Instance-aware selective branching network. CoRR abs\/1905.04849 (2019)."},{"key":"e_1_3_1_6_2","doi-asserted-by":"crossref","unstructured":"Zexuan Chang Yongshi Wang Hao Li and Zhigang Wang. 2019. Complex CNN-based equalization for communication signal. In 2019 IEEE 4th International Conference on Signal and Image Processing (ICSIP\u201919). 513\u2013517.","DOI":"10.1109\/SIPROCESS.2019.8868708"},{"key":"e_1_3_1_7_2","doi-asserted-by":"publisher","unstructured":"S. Dave Y. Kim S. Avancha S. Lee and A. Shrivastava. 2019. DMazeRunner: Executing perfectly nested loops on dataflow accelerators. ACM Trans. Embed. Comput. Syst. 18 5s (Oct. 2019) 27 pages. DOI:10.1145\/3358198","DOI":"10.1145\/3358198"},{"key":"e_1_3_1_8_2","unstructured":"David Eigen Marc\u2019Aurelio Ranzato and Ilya Sutskever. 2014. Learning factored representations in a deep mixture of experts."},{"key":"e_1_3_1_9_2","unstructured":"William Fedus Barret Zoph and Noam Shazeer. 2022. Switch transformers: Scaling to trillion parameter models with simple and efficient sparsity. arXiv:2101.03961. Retrieved from https:\/\/arxiv.org\/abs\/2101.03961 (2022)."},{"key":"e_1_3_1_10_2","doi-asserted-by":"publisher","DOI":"10.23919\/DATE56975.2023.10136926"},{"key":"e_1_3_1_11_2","unstructured":"Y. Han G. Huang S. Song L. Yang H. Wang and Y. Wang. 2021. Dynamic neural networks: A survey. CoRR abs\/2102.04906 (2021)."},{"key":"e_1_3_1_12_2","doi-asserted-by":"publisher","unstructured":"Hegde Tsai Huang Chandra Parashar and Fletcher. 2021. Mind mappings: Enabling efficient algorithm-accelerator mapping space search. CoRR abs\/2103.01489 (2021) 943\u2013958. DOI:10.1145\/3445814.3446762","DOI":"10.1145\/3445814.3446762"},{"key":"e_1_3_1_13_2","doi-asserted-by":"crossref","unstructured":"Mikko Honkala Dani Korpi and Janne M. J. Huttunen. 2021. DeepRx: Fully convolutional DL receiver. IEEE Transactions on Wireless Communications 20 6 (2021) 3925\u20133940.","DOI":"10.1109\/TWC.2021.3054520"},{"key":"e_1_3_1_14_2","unstructured":"Hoydis Cammerer Aoudia Vem Binder Marcus and Keller. 2022. Sionna: An open-source library for next-generation physical layer research. arXiv:2203.11854. Retrieved from https:\/\/arxiv.org\/abs\/2203.11854"},{"key":"e_1_3_1_15_2","doi-asserted-by":"crossref","unstructured":"Q. Huang M. Kang G. Dinh T. Norell A. Kalaiah J. Demmel J. Wawrzynek and Y. S. Shao. 2021. CoSA: Scheduling by constrained optimization for spatial accelerators. CoRR abs\/2105.01898 (2021).","DOI":"10.1109\/ISCA52012.2021.00050"},{"key":"e_1_3_1_16_2","doi-asserted-by":"publisher","DOI":"10.1162\/neco.1991.3.1.79"},{"key":"e_1_3_1_17_2","volume-title":"ICCAD","year":"2020","unstructured":"Kao and Krishna. 2020. GAMMA: Automating the HW mapping of DNN models on accelerators via genetic algorithm. In ICCAD."},{"key":"e_1_3_1_18_2","article-title":"The path to 6G","author":"Luca Stefano De","year":"2024","unstructured":"Stefano De Lucaet al.2024. The path to 6G. arXiv preprint.","journal-title":"arXiv preprint"},{"key":"e_1_3_1_19_2","doi-asserted-by":"publisher","DOI":"10.1145\/3219819.3220007"},{"key":"e_1_3_1_20_2","doi-asserted-by":"crossref","unstructured":"L. Mei P. Houshmand V. Jain S. Giraldo and M. Verhelst. 2021. ZigZag: Enlarging joint architecture-mapping design space exploration for DNN accelerators. IEEE Transactions on Computers 70 8 (2021) 1160\u20131174.","DOI":"10.1109\/TC.2021.3059962"},{"key":"e_1_3_1_21_2","doi-asserted-by":"publisher","DOI":"10.1109\/ISPASS.2019.00042"},{"key":"e_1_3_1_22_2","unstructured":"Noam Shazeer Azalia Mirhoseini Krzysztof Maziarz Andy Davis Quoc Le Geoffrey Hinton and Jeff Dean. 2017. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. CoRR abs\/1701.06538 (2017)."},{"key":"e_1_3_1_23_2","doi-asserted-by":"crossref","unstructured":"Ori Shental and Jakob Hoydis. 2019. \u201cMachine LLRning\u201d: Learning to Softly Demodulate. CoRR abs\/1907.01512 (2019).","DOI":"10.1109\/GCWkshps45667.2019.9024433"},{"key":"e_1_3_1_24_2","doi-asserted-by":"crossref","unstructured":"Ori Shental and Jakob Hoydis. 2019. \u201cMachine LLRning\u201d: Learning to softly demodulate. CoRR abs\/1907.01512 (2019) 1\u20137.","DOI":"10.1109\/GCWkshps45667.2019.9024433"},{"key":"e_1_3_1_25_2","doi-asserted-by":"crossref","unstructured":"Bram van Bolderik Souradip Sarkar Vlado Menkovski Sonia Heemstra and Manil Dev Gomony. 2024. Agile design-space exploration of dynamic layer-skipping in neural receivers. In 2024 27th Euromicro Conference on Digital System Design (DSD\u201924). 122\u2013128.","DOI":"10.1109\/DSD64264.2024.00025"},{"key":"e_1_3_1_26_2","doi-asserted-by":"crossref","unstructured":"Bram van Bolderik Vlado Menkovski Sonia Heemstra and Manil Dev Gomony. 2024. MEAN: Mixture-of-experts based neural receiver. In 2024 IFIP\/IEEE 32nd International Conference on Very Large Scale Integration (VLSI-SoC24). 1\u20134.","DOI":"10.1109\/VLSI-SoC62099.2024.10767787"},{"key":"e_1_3_1_27_2","unstructured":"Andreas Veit and Serge Belongie. 2017. Convolutional networks with adaptive computation graphs. CoRR abs\/1711.11503 (2017)."},{"key":"e_1_3_1_28_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM.2016.22"},{"key":"e_1_3_1_29_2","unstructured":"Xin Wang Fisher Yu Zi-Yi Dou Trevor Darrell and Joseph E. Gonzalez. 2017. SkipNet: Learning Dynamic Routing in Convolutional Networks. CoRR abs\/1711.09485 (2017)."},{"key":"e_1_3_1_30_2","unstructured":"Xin Wang Fisher Yu Ruth Wang Yi-An Ma Azalia Mirhoseini Trevor Darrell and Joseph E. Gonzalez. 2018. Deep mixture of experts via shallow embedding. CoRR abs\/1806.01531 (2018)."},{"key":"e_1_3_1_31_2","unstructured":"Paul N. Whatmough Chuteng Zhou Patrick Hansen Shreyas Kolala Venkataramanaiah Jae-sun Seo and Matthew Mattina. 2019. FixyNN: Efficient hardware for mobile computer vision via transfer learning."},{"key":"e_1_3_1_32_2","unstructured":"N. Wu H. Yang Y. Xie P. Li and C. Hao. 2022. High-level synthesis performance prediction using GNNs: Benchmarking modeling and advancingCoRR abs\/2201.06848 (2022)."},{"key":"e_1_3_1_33_2","doi-asserted-by":"crossref","unstructured":"X. Yang M. Gao Q. Liu J. Setter J. Pu A. Nayak S. Bell K. Cao H. Ha P. Raina C. Kozyrakis and M. Horowitz. 2020. Interstellar: Using Halide\u2019s scheduling language to analyze DNN accelerators. In Proceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems ASPLOS\u201920 ACM 369\u2013383.","DOI":"10.1145\/3373376.3378514"},{"key":"e_1_3_1_34_2","unstructured":"Penghang Yin Jiancheng Lyu Shuai Zhang Stanley Osher Yingyong Qi and Jack Xin. 2019. Understanding straight-through estimator in training activation quantized neural nets. CoRR abs\/1903.05662 (2019)."},{"key":"e_1_3_1_35_2","article-title":"DemodNet: Learning soft demodulation from hard information using CNN","author":"Zheng Shilian","year":"2022","unstructured":"Shilian Zheng, Xiaoyu Zhou, Shichuan Chen, Peihan Qi, and Xiaoniu Yang. 2022. DemodNet: Learning soft demodulation from hard information using CNN. IEEE ICC arXiv:2011.11337. Retrieved from https:\/\/arxiv.org\/abs\/2011.11337","journal-title":"IEEE ICC"}],"container-title":["ACM Transactions on Embedded Computing Systems"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3765519","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,13]],"date-time":"2026-05-13T13:58:40Z","timestamp":1778680720000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3765519"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,13]]},"references-count":34,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2026,5,31]]}},"alternative-id":["10.1145\/3765519"],"URL":"https:\/\/doi.org\/10.1145\/3765519","relation":{},"ISSN":["1539-9087","1558-3465"],"issn-type":[{"value":"1539-9087","type":"print"},{"value":"1558-3465","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,13]]},"assertion":[{"value":"2024-12-01","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-07-20","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-05-13","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}