<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE root>
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:ali="http://www.niso.org/schemas/ali/1.0/" article-type="research-article" dtd-version="1.2" xml:lang="en"><front><journal-meta><journal-id journal-id-type="publisher-id">Informacionnye Tehnologii</journal-id><journal-title-group><journal-title xml:lang="en">Informacionnye Tehnologii</journal-title><trans-title-group xml:lang="ru"><trans-title>Информационные технологии</trans-title></trans-title-group></journal-title-group><issn publication-format="print">1684-6400</issn><publisher><publisher-name xml:lang="en">New Technologies Publishing House</publisher-name></publisher></journal-meta><article-meta><article-id pub-id-type="publisher-id">702282</article-id><article-id pub-id-type="doi">10.17587/it.31.364-369</article-id><article-categories><subj-group subj-group-type="toc-heading" xml:lang="en"><subject>Neural network technologies</subject></subj-group><subj-group subj-group-type="toc-heading" xml:lang="ru"><subject>Нейросетевые технологии</subject></subj-group><subj-group subj-group-type="article-type"><subject>Research Article</subject></subj-group></article-categories><title-group><article-title xml:lang="en">Application of integer tables for quantisation of activation functions of neural networks</article-title><trans-title-group xml:lang="ru"><trans-title>Применение целочисленных таблиц для квантизации функций активаций нейронных сетей</trans-title></trans-title-group></title-group><contrib-group><contrib contrib-type="author"><name-alternatives><name xml:lang="en"><surname>Vasilev</surname><given-names>А. А.</given-names></name><name xml:lang="ru"><surname>Васильев</surname><given-names>А. А.</given-names></name></name-alternatives><address><country country="RU">Russian Federation</country></address><bio xml:lang="en"><p>Middle Engineer</p></bio><bio xml:lang="ru"><p>ст. инженер</p></bio><email>artvasilev@alphachip.ru</email><xref ref-type="aff" rid="aff1"/></contrib><contrib contrib-type="author"><name-alternatives><name xml:lang="en"><surname>Kapitanov</surname><given-names>А. I.</given-names></name><name xml:lang="ru"><surname>Капитанов</surname><given-names>А. И.</given-names></name></name-alternatives><address><country country="RU">Russian Federation</country></address><bio xml:lang="en"><p>Associate Professor</p></bio><bio xml:lang="ru"><p>доц.</p></bio><email>andrey@kapdx.ru</email><xref ref-type="aff" rid="aff2"/></contrib></contrib-group><aff-alternatives id="aff1"><aff><institution xml:lang="en">LLC "Alphachip"</institution></aff><aff><institution xml:lang="ru">ООО "Альфачип"</institution></aff></aff-alternatives><aff-alternatives id="aff2"><aff><institution xml:lang="en">SPINTech Institute, National Research University "MIET"</institution></aff><aff><institution xml:lang="ru">Институт СПИНТех, Национальный исследовательский университет "МИЭТ"</institution></aff></aff-alternatives><pub-date date-type="pub" iso-8601-date="2025-07-15" publication-format="electronic"><day>15</day><month>07</month><year>2025</year></pub-date><volume>31</volume><issue>7</issue><issue-title xml:lang="en"/><issue-title xml:lang="ru"/><fpage>364</fpage><lpage>369</lpage><history><date date-type="received" iso-8601-date="2026-02-06"><day>06</day><month>02</month><year>2026</year></date><date date-type="accepted" iso-8601-date="2026-02-06"><day>06</day><month>02</month><year>2026</year></date></history><permissions><copyright-statement xml:lang="en">Copyright ©; 2025, Informacionnye Tehnologii</copyright-statement><copyright-statement xml:lang="ru">Copyright ©; 2025, Информационные технологии</copyright-statement><copyright-year>2025</copyright-year><copyright-holder xml:lang="en">Informacionnye Tehnologii</copyright-holder><copyright-holder xml:lang="ru">Информационные технологии</copyright-holder></permissions><self-uri xlink:href="https://journals.eco-vector.com/1684-6400/article/view/702282">https://journals.eco-vector.com/1684-6400/article/view/702282</self-uri><abstract xml:lang="en"><p>The paper considers the problem of efficient hardware implementation of nonlinear activation functions of neural networks under low-bit computing conditions. Standard activations, such as sigmoid and hyperbolic tangent, require resource-intensive floating-point operations, which limits their use on microcontrollers, FPGAs and other peripheral platforms. As a solution, an approach based on precomputed integer substitution tables (LUTs) is proposed to reduce computational complexity and power consumption. Using the example of the SiLU activation function widely used in popular object detection networks (e.g., YOLO), the quantisation procedure is demonstrated, the principles of constructing and using LUTs are formulated, and a practical algorithm for computing activations using them is described.</p></abstract><trans-abstract xml:lang="ru"><p>Рассматривается проблема эффективной аппаратной реализации нелинейных функций активации нейрон­ных сетей в условиях низкоразрядных вычислений. Стандартные функции активации, такие как сигмоида и гиперболический тангенс, требуют ресурсоемких операций с плавающей запятой, что ограничивает их использование на микроконтроллерах, FPGA и других периферийных платформах. В качестве решения предложен подход на основе предварительно рассчитанных целочисленных таблиц подстановки (LUT), позволяющий со­кратить вычислительную сложность и потребление энергии. На примере функции активации SiLU, широко применяемой в популярных сетях детектирования объектов (например, YOLO), продемонстрирована процедура квантизации, сформулированы принципы построения и использования LUT, а также описан практический ал­горитм вычисления функций активаций с их помощью.</p></trans-abstract><kwd-group xml:lang="en"><kwd>quantisation</kwd><kwd>integrated circuits</kwd><kwd>neural networks</kwd><kwd>convolutional neural networks</kwd><kwd>hardware implementation of neural networks</kwd></kwd-group><kwd-group xml:lang="ru"><kwd>квантизация</kwd><kwd>интегральные схемы</kwd><kwd>нейронные сети</kwd><kwd>сверточные нейронные сети</kwd><kwd>аппаратная реализация нейронных сетей</kwd></kwd-group><funding-group/></article-meta></front><body></body><back><ref-list><ref id="B1"><label>1.</label><mixed-citation>Romanov A. Y., Stempkovsky A. L., Lariushkin I. V., Novoselov G. E., Solovyev R. A., Starykh V. A., Romanova I. I., Telpukhov D. V., Mkrtchan I. A. Analysis of posit and bfloat arithmetic of real numbers for machine learning // IEEE Access. 2021. N. 9. P. 82318—82324.</mixed-citation></ref><ref id="B2"><label>2.</label><mixed-citation>Zhao X., Wang Y., Cai X., Liu C., Zhang L. Linear symmetric quantization of neural networks for low-precision integer hardware //International Conference on Learning Representations. 2020, April.</mixed-citation></ref><ref id="B3"><label>3.</label><mixed-citation>Krishnamoorthi R. Quantizing deep convolutional networks for efficient inference: A whitepaper // arXiv preprint arXiv:1806.08342. 2018.</mixed-citation></ref><ref id="B4"><label>4.</label><mixed-citation>Solovyev R. A., Kalinin A. A., Kustov A. G., Telpukhov D. V., Ruhlov V. S. FPGA implementation of convolutional neural networks with fixed-point calculations // CoRR. 2018. [Online]. Available: https://openreview.net/forum?id=ktBJlHahoT.</mixed-citation></ref><ref id="B5"><label>5.</label><mixed-citation>Solovyev R. A., Telpukhov D. V., Romanova I. I., Kustov A. G., Mkrtchan I. A. Real-time Object Detection with FPGA Using CenterNet // 2021 IEEE Conference of Russian Young Researchers in Electrical and Electronic Engineering (ElConRus). 2021, January. P. 2029—2034.</mixed-citation></ref><ref id="B6"><label>6.</label><mixed-citation>Nagel M., Fournarakis M., Amjad R. A., Bondarenko Y., Van Baalen M., Blankevoort T. A white paper on neural network quantization // arXiv preprint arXiv:2106.08295. 2021.</mixed-citation></ref><ref id="B7"><label>7.</label><mixed-citation>Yang J., Shen X., Xing J., Tian X., Li H., Deng B., Huang J., Hua X. S. Quantization networks //Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019. P. 7308—7316.</mixed-citation></ref><ref id="B8"><label>8.</label><mixed-citation>Gholami A., Kim S., Dong Z., Yao Z., Mahoney M. W., Keutzer K. A survey of quantization methods for efficient neural network inference // Low-power computer vision. Chapman and Hall/CRC. 2022. P. 291—326.</mixed-citation></ref><ref id="B9"><label>9.</label><mixed-citation>Hubara I., Courbariaux M., Soudry D., El-Yaniv R., Bengio Y. Quantized neural networks: Training neural networks with low precision weights and activations // Journal of machine learning research. 2018. Vol. 18, N. 187. P. 1—30.</mixed-citation></ref><ref id="B10"><label>10.</label><mixed-citation>Hagiescu A., Langhammer M., Pasca B., Colangelo P., Thong J., Ilkhani N. Bfloat MLP training accelerator for FPGAs // 2019 International Conference on ReConFigurable Computing and FPGAs (ReConFig). 2019, December. P. 1—5.</mixed-citation></ref><ref id="B11"><label>11.</label><mixed-citation>Zhai H. Z., Du J. W., Ai Y. H., Hu T. J. Edge Deployment of Deep Networks for Visual Object Detection: A Review // IEEE Sensors Journal. 2024. [Online]. Available: https://ieeexplore.ieee. org/abstract/document/10786287</mixed-citation></ref></ref-list></back></article>
