Modeling of systems and processes Modeling of systems and processes Моделирование систем и процессов 2219-0767 120340 10.12737/2219-0767-2026-19-1-94-105 Технические науки Технические науки Multi-criteria evolutionary design of neural network architectures for decision-making based on tabular data under uncertainty Многокритериальное эволюционное проектирование нейросетевых архитектур для принятия решений на основе табличных данных в условиях неопределенности Скворцов Александр Александрович Skvortsov Alexander Alexandrovich Соловьев Денис Сергеевич Solovjev Denis Sergeevich solovjevdenis@mail.ru Арапов Денис Владимирович Arapov Denis Vladimirovich Александров Денис Александрович Aleksandrov Denis Aleksandrovich Тамбовский государственный университет имени Г. Р. Державина Россия Derzhavin Tambov State University Russian Federation 24 04 2026 24 04 2026 19 1 94 105 08 04 2026 https://naukaru.ru/en/nauka/article/120340/view

В статье рассматривается задача системного проектирования архитектур нейронных сетей для обработки табличных данных в условиях жестких ресурсных ограничений, характерных для встраиваемых и периферийных систем. Целью работы является разработка методики, формализующей данную задачу как проблему многокритериальной условной оптимизации в дискретном пространстве гиперпараметров и рассматривающей процесс синтеза как поддержку принятия решений в условиях неопределенности. Предложена методика многокритериального эволюционного синтеза, основанная на алгоритме Multi-Island Genetic Algorithm (MIGA), который интегрирует островную модель эволюции для поддержания разнообразия и механизм отбора NSGA-II для построения аппроксимации Парето-фронта. В качестве конфликтующих критериев оптимизации использованы точность классификации, объем памяти, требуемый для хранения модели, и латентность инференса. Для экспериментальной валидации методики выбраны три публичных набора табличных данных, репрезентирующих различные прикладные сценарии и уровни сложности. Разработан и реализован программный комплекс с трехуровневой архитектурой, обеспечивающий полный цикл автоматизированного проектирования – от адаптивного анализа данных до визуализации результатов. Проведенный сравнительный анализ с базовыми методами (логистическая регрессия, решающее дерево, градиентный бустинг) показал, что предложенная методика позволяет синтезировать модели, которые при сопоставимой точности на порядки компактнее и быстрее моделей градиентного бустинга, а в случае сложных нелинейных зависимостей при малой выборке – статистически значимо превосходят их по точности. Результаты работы подтверждают практическую значимость методики для снижения трудоемкости проектирования и предоставления разработчику множества количественно обоснованных компромиссных решений, адекватных заданным аппаратным ограничениям.

This paper addresses the systematic design of neural network architectures for tabular data processing under stringent resource constraints typical for embedded and edge systems. The work aims to develop a methodology that formalizes this task as a constrained multi-objective optimization problem within a discrete hyperparameter space, framing the synthesis process as decision support under uncertainty. A methodology for multi-criteria evolutionary synthesis is proposed, based on the Multi-Island Genetic Algorithm (MIGA), which integrates an island model of evolution to maintain population diversity and the NSGA-II selection mechanism to construct a Pareto front approximation. Conflicting optimization criteria include classification accuracy, memory footprint required for model storage, and inference latency. For the experimental validation of the methodology, three public tabular datasets representing different application scenarios and complexity levels were selected. A software framework with a three-tier architecture was developed and implemented, supporting the full cycle of automated design—from adaptive data analysis to results visualization. A comparative analysis with baseline methods (logistic regression, decision tree, gradient boosting) demonstrated that the proposed methodology can synthesize models that, with comparable accuracy, are orders of magnitude more compact and faster than gradient boosting models. In cases involving complex nonlinear dependencies with small sample sizes, the synthesized models statistically significantly outperform the baselines in accuracy. The results confirm the practical significance of the methodology for reducing design complexity and providing developers with a set of quantitatively justified trade-off solutions that comply with given hardware constraints.

 Многокритериальная оптимизация эволюционный синтез архитектур нейронные сети табличные данные ресурсо-ограниченные системы многослойный персептрон системный анализ принятие решений Парето-оптимальность автоматизированное проектирование Multi-objective optimization evolutionary architecture synthesis neural networks tabular data resource-constrained systems multilayer perceptron systems analysis decision-making Pareto optimality automated design

Elsken T., Metzen J. H., Hutter F. Neural architecture search: A survey //Journal of Machine Learning Research. – 2019. – V. 20. – №. 55. – P. 1-21. Elsken T., Metzen J. H., Hutter F. Neural architecture search: A survey //Journal of Machine Learning Research. – 2019. – V. 20. – №. 55. – P. 1-21. Chen T., Guestrin C. XGBoost: A Scalable Tree Boosting System // Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD '16). – 2016. – P. 785–794. DOI: 10.1145/2939672.293978 Chen T., Guestrin C. XGBoost: A Scalable Tree Boosting System // Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD '16). – 2016. – P. 785–794. DOI: 10.1145/2939672.293978 Grinsztajn L., Oyallon E., Varoquaux G. Why do tree-based models still outperform deep learning on typical tabular data? //Advances in neural information processing systems. – 2022. – V. 35. – P. 507-520. Grinsztajn L., Oyallon E., Varoquaux G. Why do tree-based models still outperform deep learning on typical tabular data? //Advances in neural information processing systems. – 2022. – V. 35. – P. 507-520. Dorogush A.V., Ershov V., Gulin A. CatBoost: gradient boosting with categorical features support // arXiv:1810,11363 [cs.LG]. – 2018. DOI: 10.48550/arXiv.1810.11363 Dorogush A.V., Ershov V., Gulin A. CatBoost: gradient boosting with categorical features support // arXiv:1810,11363 [cs.LG]. – 2018. DOI: 10.48550/arXiv.1810.11363 Prokhorenkova L. et al. CatBoost: unbiased boosting with categorical features // Advances in Neural Information Processing Systems. – 2018. – V. 31. – P. 6637–6647. Prokhorenkova L. et al. CatBoost: unbiased boosting with categorical features // Advances in Neural Information Processing Systems. – 2018. – V. 31. – P. 6637–6647. Banbury C. et al. MLPerf Tiny Benchmark // arXiv:2106,07597 [cs.LG]. – 2021. DOI: 10.48550/arXiv.2106.07597 Banbury C. et al. MLPerf Tiny Benchmark // arXiv:2106,07597 [cs.LG]. – 2021. DOI: 10.48550/arXiv.2106.07597 Sarker Z., Mitra S., Gandhi S. A Comprehensive Review on Compiler in Neural Networks //2023 IEEE International Conference on Contemporary Computing and Communications (InC4). – IEEE, 2023. – V. 1. – P. 1-6. DOI: 10.1109/InC457730.2023.10263112 Sarker Z., Mitra S., Gandhi S. A Comprehensive Review on Compiler in Neural Networks //2023 IEEE International Conference on Contemporary Computing and Communications (InC4). – IEEE, 2023. – V. 1. – P. 1-6. DOI: 10.1109/InC457730.2023.10263112 Айвазян С. А. Прикладная статистика. Основы эконометрики.-М.: ЮНИТИ-ДАНА. – 2001. – 432 с. Aivazyan S. A. Applied Statistics. Fundamentals of Econometrics. – Moscow: UNITY-DANA, 2001. – 432 p. Gorishniy Y. et al. Revisiting deep learning models for tabular data //Advances in neural information processing systems. – 2021. – V. 34. – P. 18932-18943. Gorishniy Y. et al. Revisiting deep learning models for tabular data //Advances in neural information processing systems. – 2021. – V. 34. – P. 18932-18943. Часовских А. А. и др. Автоматизация поиска архитектур искусственных нейронных сетей //Интеллектуальные системы. Теория и приложения. – 2023. – Т. 27. – №. 4. – С. 5-27. Chasovskikh A. A. et al. Automation of Neural Network Architecture Search // Intelligent Systems. Theory and Applications. – 2023. – Vol. 27. – No. 4. – P. 5–27. Feurer M., Hutter F. Hyperparameter Optimization // In: Hutter F., Kotthoff L., Vanschoren J. (eds) Automated Machine Learning: Methods, Systems, Challenges. The Springer Series on Challenges in Machine Learning. – Springer, Cham. – 2019. – P. 3–38. Feurer M., Hutter F. Hyperparameter Optimization // In: Hutter F., Kotthoff L., Vanschoren J. (eds) Automated Machine Learning: Methods, Systems, Challenges. The Springer Series on Challenges in Machine Learning. – Springer, Cham. – 2019. – P. 3–38. Warden P., Situnayake D. TinyML: Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-Power Microcontrollers. – Sebastopol, CA: O'Reilly Media. – 2020. – 504 p. Warden P., Situnayake D. TinyML: Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-Power Microcontrollers. – Sebastopol, CA: O'Reilly Media. – 2020. – 504 p. Liu H., Simonyan K., Yang Y. Darts: Differentiable architecture search //arXiv preprint arXiv:1806,09055. – 2018. DOI: 10.48550/arXiv.1806.09055 Liu H., Simonyan K., Yang Y. Darts: Differentiable architecture search //arXiv preprint arXiv:1806,09055. – 2018. DOI: 10.48550/arXiv.1806.09055 Xing N. et al. Anytime neural architecture search on tabular data //arXiv preprint arXiv:2403,10318. – 2024. DOI: 10.48550/arXiv.2403.10318 Xing N. et al. Anytime neural architecture search on tabular data //arXiv preprint arXiv:2403,10318. – 2024. DOI: 10.48550/arXiv.2403.10318 Deb K. et al. A fast and elitist multiobjective genetic algorithm: NSGA-II //IEEE transactions on evolutionary computation. – 2002. – V. 6. – №. 2. – P. 182-197. DOI: 10.1109/4235.996017 Deb K. et al. A fast and elitist multiobjective genetic algorithm: NSGA-II //IEEE transactions on evolutionary computation. – 2002. – V. 6. – №. 2. – P. 182-197. DOI: 10.1109/4235.996017 He X., Zhao K., Chu X. AutoML: A Survey of the State-of-the-Art // Knowledge-Based Systems. – 2021. – V. 212. – P. 106622. DOI: 10.1016/j.knosys.2020.106622 He X., Zhao K., Chu X. AutoML: A Survey of the State-of-the-Art // Knowledge-Based Systems. – 2021. – V. 212. – P. 106622. DOI: 10.1016/j.knosys.2020.106622 Hutter F., Kotthoff L., Vanschoren J. (Eds.) Automated Machine Learning: Methods, Systems, Challenges. – Springer, 2019. – 223 p. Hutter F., Kotthoff L., Vanschoren J. (Eds.) Automated Machine Learning: Methods, Systems, Challenges. – Springer, 2019. – 223 p. Воронин А., Зиатдинов Ю., Антонюк А. Многокритериальная оптимизация архитектуры нейросетевых классификаторов //Classification, Forecasting, Data Mining. – 2009. – С. 32–39. Voronin A., Ziatdinov Yu., Antonyuk A. Multi-criteria Optimization of Neural Network Classifier Architectures // Classification, Forecasting, Data Mining. – 2009. – P. 32–39. Arik S. Ö., Pfister T. Tabnet: Attentive interpretable tabular learning //Proceedings of the AAAI conference on artificial intelligence. – 2021. – V. 35. – №. 8. – P. 6679-6687. DOI: 10.1609/aaai.v35i8.16826 Arik S. Ö., Pfister T. Tabnet: Attentive interpretable tabular learning //Proceedings of the AAAI conference on artificial intelligence. – 2021. – V. 35. – №. 8. – P. 6679-6687. DOI: 10.1609/aaai.v35i8.16826 Gorishniy Y. et al. Revisiting deep learning models for tabular data //Advances in neural information processing systems. – 2021. – V. 34. – P. 18932-18943. Gorishniy Y. et al. Revisiting deep learning models for tabular data //Advances in neural information processing systems. – 2021. – V. 34. – P. 18932-18943. He Y. et al. Filter pruning via geometric median for deep convolutional neural networks acceleration //Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. – 2019. – P. 4340-4349. He Y. et al. Filter pruning via geometric median for deep convolutional neural networks acceleration //Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. – 2019. – P. 4340-4349. Jacob B. et al. Quantization and training of neural networks for efficient integer-arithmetic-only inference //Proceedings of the IEEE conference on computer vision and pattern recognition. – 2018. – P. 2704-2713. Jacob B. et al. Quantization and training of neural networks for efficient integer-arithmetic-only inference //Proceedings of the IEEE conference on computer vision and pattern recognition. – 2018. – P. 2704-2713. Скворцов А.А., Анурьева М.С., Солодовников А.Н. Интеллектуальная система классификации текстов в условиях лингвистической неопределенности // Программная инженерия. – 2025. – Т. 16. – №. 11. – С. 583–593. DOI: 10.17587/prin.16.583-593. Skvortsov A. A., Anuryeva M. S., Solodovnikov A. N. Intelligent Text Classification System under Linguistic Uncertainty // Software Engineering. – 2025. – Vol. 16. – No. 11. – P. 583–593. – DOI: 10.17587/prin.16.583-593. Whitley D., Rana S., Heckendorn R. B. The island model genetic algorithm: On separability, population size and convergence //Journal of computing and information technology. – 1999. – V. 7. – №. 1. – P. 33-47. Whitley D., Rana S., Heckendorn R. B. The island model genetic algorithm: On separability, population size and convergence //Journal of computing and information technology. – 1999. – V. 7. – №. 1. – P. 33-47. Lu Z. et al. Nsga-net: neural architecture search using multi-objective genetic algorithm // Proceedings of the genetic and evolutionary computation conference. – 2019. – P. 419-427. DOI: 10.1145/3321707.332172 Lu Z. et al. Nsga-net: neural architecture search using multi-objective genetic algorithm // Proceedings of the genetic and evolutionary computation conference. – 2019. – P. 419-427. DOI: 10.1145/3321707.332172 Deb K. et al. A fast and elitist multiobjective genetic algorithm: NSGA-II //IEEE transactions on evolutionary computation. – 2002. – V. 6. – №. 2. – P. 182-197. DOI: 10.1109/4235.996017 Deb K. et al. A fast and elitist multiobjective genetic algorithm: NSGA-II //IEEE transactions on evolutionary computation. – 2002. – V. 6. – №. 2. – P. 182-197. DOI: 10.1109/4235.996017 IoT Occupancy Detection Dataset [Электронный ресурс]. – UCI Machine Learning Repository. – URL: https://archive.ics.uci.edu/ml/datasets/occupancy+detection (дата обращения: 20.11.2025). IoT Occupancy Detection Dataset [Electronic resource]. – UCI Machine Learning Repository. – URL: https://archive.ics.uci.edu/ml/datasets/occupancy+detection (date of request: 20.11.2025). Airline Passenger Satisfaction Dataset [Электронный ресурс]. – Kaggle. – URL: https://www.kaggle.com/datasets/teejmahal20/airline-passenger-satisfaction (дата обращения: 20.11.2025). Airline Passenger Satisfaction Dataset [Electronic resource]. – Kaggle. – URL: https://www.kaggle.com/datasets/teejmahal20/airline-passenger-satisfaction (date of request: 20.11.2025). Wine Quality (Red Wine) Dataset [Электронный ресурс]. – UCI Machine Learning Repository. – URL: https://archive.ics.uci.edu/ml/datasets/wine+quality (дата обращения: 20.11.2025). Wine Quality (Red Wine) Dataset [Electronic resource]. – UCI Machine Learning Repository. – URL: https://archive.ics.uci.edu/ml/datasets/wine+quality (date of request: 20.11.2025).