bsuirДоклады БГУИРDoklady BGUIR1729-76482708-0382БГУИР10.35596/1729-7648-2026-24-4-38-46bsuir-4402Research ArticleСтатьиРазработка платформенно-ориентированного алгоритма синтеза оптимальной нейросетевой архитектуры для мониторинга лазерной полировки кварцевых стеколDevelopment of a Platform-Oriented Algorithm for Synthesizing an Optimal Neural Network Architecture for Monitoring Laser Polishing of Quartz GlassПрохоренкоВ. А.ProkhorenkoV.

Прохоренко Владислав Александрович, ст. преп. каф. математических проблем управления и информатики

246028, Гомель, ул. Кирова, 119

Тел.: +375 232 51-03-04

Prokhorenko V., Senior Lecturer at the Department of Mathematical Problems of Control and Computer Science

246028, Gomel, Kirova St., 119

shnysct@gmail.comСмородинВ. С.SmorodinV.

Смородин В. С., д-р техн. наук, проф., зав. каф. математических проблем управления и информатики

246028, Гомель, ул. Кирова, 119

Smorodin V., Dr. Sci. (Tech.), Professor, Head of the Department of Mathematical Problems of Control and Computer Science

246028, Gomel, Kirova St., 119

НикитюкЮ. В.NikitjukYu.

Никитюк Ю. В., канд. физ.-мат. наук, доц., проректор по учебной работе

246028, Гомель, ул. Кирова, 119

Nikitjuk Yu., Cand. Sci. (Phys. and Math.), Associate Professor, Vice-Rector for Academic Affairs

246028, Gomel, Kirova St., 119

Гомельский государственный университет имени Франциска СкориныБеларусьFrancisk Skorina Gomel State UniversityBelarus2026300820262443846Copyright © Прохоренко В.А., Смородин В.С., Никитюк Ю.В., 20262026Прохоренко В.А., Смородин В.С., Никитюк Ю.В.Prokhorenko V., Smorodin V., Nikitjuk Y.Данная работа распространяется под лицензией Creative Commons Attribution 4.0.This work is licensed under a Creative Commons Attribution 4.0 License.https://doklady.bsuir.by/jour/article/view/4402

Предложен алгоритм эволюционного синтеза архитектур нейронных сетей, ориентированный на решение задач мониторинга технологических процессов в условиях ограниченных вычислительных ресурсов и обучающих выборок в режиме реального времени. Алгоритм основан на модифицированном подходе нейроэволюции, в котором геном особи включает в себя граф крупных функциональных блоков и происходит отслеживание истории инноваций. Оценка особей популяции выполнялась многокритери- альным алгоритмом NSGA-II на основе точности классификации и времени прямого прохода сети на целевом устройстве. Работоспособность алгоритма продемонстрирована на задаче классификации изображений с использованием набора данных CIFAR-10, а также на прикладной задаче классификации режимов работы при реализации технологической операции лазерной однолучевой полировки кварцевых стекол.

An evolutionary synthesis algorithm for neural network architectures is proposed. It is designed to address real-time process monitoring tasks with limited computing resources and training data. The algorithm is based on a modified neuroevolution approach, in which the individual genome comprises a graph of large functional blocks and the history of innovations is tracked. The evaluation on population individuals was performed by the multi-criteria algorithm NSGA-II based on classification accuracy and the forward pass time network on the target device. The algorithm’s performance is demonstrated on an image classification task using the CIFAR-10 dataset, as well as on the applied task of classifying operating modes during the single-beam laser polishing of quartz glass.

нейронные сетинейроэволюцияпоиск нейросетевых архитектурлазерная обработка материаловмониторинг технологических процессовneural networksneuroevolutionsearch for neural network architectureslaser processing of materialsmonitoring of technological processesReferencesStanley K. O., Miikkulainen R. (2002) Evolving Neural Networks through Augmenting Topologies. Evolutionary Computation. 10 (2), 99–127.Stanley K. O., Miikkulainen R. (2002) Evolving Neural Networks through Augmenting Topologies. Evolutionary Computation. 10 (2), 99–127.Miikkulainen R., Liang J., Meyerson E., Rawal A., Fink D., Francon O. (2017) Evolving Deep Neural Networks. arXiv preprint arXiv:1703.00548.Miikkulainen R., Liang J., Meyerson E., Rawal A., Fink D., Francon O. (2017) Evolving Deep Neural Networks. arXiv preprint arXiv:1703.00548.Zoph B., Le Q. V. (2016) Neural Architecture Search with Reinforcement Learning. arXiv preprint arXiv:1611.01578.Zoph B., Le Q. V. (2016) Neural Architecture Search with Reinforcement Learning. arXiv preprint arXiv:1611.01578.Elsken T., Metzen J. H., Hutter F. (2019) Neural Architecture Search: A Survey. Journal of Machine Learning Research. 20 (55), 1–21.Elsken T., Metzen J. H., Hutter F. (2019) Neural Architecture Search: A Survey. Journal of Machine Learning Research. 20 (55), 1–21.Termritthikun C., Jamtsho Y., Ieamsaard J., Muneesawang P., Lee I. (2021) EEEA-Net: An Early Exit Evolutionary Neural Architecture Search. arXiv preprint arXiv:2108.06156.Termritthikun C., Jamtsho Y., Ieamsaard J., Muneesawang P., Lee I. (2021) EEEA-Net: An Early Exit Evolutionary Neural Architecture Search. arXiv preprint arXiv:2108.06156.Tan M., Chen B., Pang R., Vasudevan V., Sandler M., Howard A., et al. (2019) MnasNet: Platform-Aware Neural Architecture Search for Mobile. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Tan M., Chen B., Pang R., Vasudevan V., Sandler M., Howard A., et al. (2019) MnasNet: Platform-Aware Neural Architecture Search for Mobile. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Sandler M., Howard A., Zhu M., Zhmoginov A., Chen L.-C. (2018) MobileNetV2: Inverted Residuals and Linear Bottlenecks. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Sandler M., Howard A., Zhu M., Zhmoginov A., Chen L.-C. (2018) MobileNetV2: Inverted Residuals and Linear Bottlenecks. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Howard A., Sandler M., Chu G., Chen L., Chen B., Tan M., et al. (2019) Searching for MobileNetV3. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV).Howard A., Sandler M., Chu G., Chen L., Chen B., Tan M., et al. (2019) Searching for MobileNetV3. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV).Tan M., Le Q. V. (2021) EfficientNetV2: Smaller Models and Faster Training. Proceedings of the 38th International Conference on Machine Learning (ICML).Tan M., Le Q. V. (2021) EfficientNetV2: Smaller Models and Faster Training. Proceedings of the 38th International Conference on Machine Learning (ICML).Chen Y., Dai X., Liu M., Chen D., Yuan L., Liu Z. (2020) Dynamic Convolution: Attention over Convolution Kernels. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Chen Y., Dai X., Liu M., Chen D., Yuan L., Liu Z. (2020) Dynamic Convolution: Attention over Convolution Kernels. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Ma N., Zhang X., Zheng H.-T., Sun J. (2018) ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design. Proceedings of the European Conference on Computer Vision (ECCV).Ma N., Zhang X., Zheng H.-T., Sun J. (2018) ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design. Proceedings of the European Conference on Computer Vision (ECCV).Liu Z., Mao H., Wu C.-Y., Feichtenhofer C., Darrell T., Xie S. (2022) A ConvNet for the 2020s. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Liu Z., Mao H., Wu C.-Y., Feichtenhofer C., Darrell T., Xie S. (2022) A ConvNet for the 2020s. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Qin D., Leichner C., Delakis M., Fornoni M., Luo S., Yang F., et al. (2024) MobileNetV4: Universal Models for the Mobile Ecosystem. arXiv preprint arXiv:2404.10518.Qin D., Leichner C., Delakis M., Fornoni M., Luo S., Yang F., et al. (2024) MobileNetV4: Universal Models for the Mobile Ecosystem. arXiv preprint arXiv:2404.10518.Dong X., Hu Y., Yang Y., Tian Y. (2021) Differentiable Architecture Search with Directional Derivatives. Advances in Neural Information Processing Systems (NeurIPS).Dong X., Hu Y., Yang Y., Tian Y. (2021) Differentiable Architecture Search with Directional Derivatives. Advances in Neural Information Processing Systems (NeurIPS).Chen T., Xu Y., Wang S., Li H. (2020) Progressive Supernet Optimization for Efficient Neural Architecture Search. Proceedings of the 37th International Conference on Machine Learning (ICML).Chen T., Xu Y., Wang S., Li H. (2020) Progressive Supernet Optimization for Efficient Neural Architecture Search. Proceedings of the 37th International Conference on Machine Learning (ICML).Cai H., Zhu L., Han S. (2020) Once for All: Train One Network and Specialize it for Efficient Deployment. International Conference on Learning Representations (ICLR).Cai H., Zhu L., Han S. (2020) Once for All: Train One Network and Specialize it for Efficient Deployment. International Conference on Learning Representations (ICLR).Krizhevsky A. (2009) Learning Multiple Layers of Features from Tiny Images.Krizhevsky A. (2009) Learning Multiple Layers of Features from Tiny Images.Emelyanov V. A., Shershnev E. B., Nikityuk Yu. V., Sokolov S. I., Aushev I. Yu. (2022) Optimization of Laser Polishing of Quartz Glass Using the Method of Computational Experiment Design. Problems of Physics, Mathematics and Engineering. 51 (2), 26–30. https://doi.org/10.54341/20778708_2022_2_51_26 (in Russian).Emelyanov V. A., Shershnev E. B., Nikityuk Yu. V., Sokolov S. I., Aushev I. Yu. (2022) Optimization of Laser Polishing of Quartz Glass Using the Method of Computational Experiment Design. Problems of Physics, Mathematics and Engineering. 51 (2), 26–30. https://doi.org/10.54341/20778708_2022_2_51_26 (in Russian).Deb K., Pratap A., Agarwal S., Meyarivan T. (2002) A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation. 6 (2), 182–197.Deb K., Pratap A., Agarwal S., Meyarivan T. (2002) A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation. 6 (2), 182–197.

The authors declare that there are no conflicts of interest present.