2687-0517 18 4 2025 1-122

RAR RUS 7-19 F-6480-2013 7004013271 0000-0002-5637-1420 Peter the Great St.Petersburg Polytechnic University Lev V. lev.utkin@mail.ru

Polytechnicheskaya, 29, St.Petersburg, Russia, 195251

Peter the Great St. Petersburg Polytechnic University Konstantinov Andrei andrue.konst@gmail.com

St. Petersburg, Russian Federation

0000-0001-8749-9470 Peter the Great St. Petersburg Polytechnic University Verbova Natalia

St. Petersburg, Russian Federation

Concept-Based Learning in Heterogeneous Treatment Effect Estimating Heterogeneous Treatment Effects (HTE) is crucial for personalized decision-making in medicine, economics and engineering. While machine learning models for Conditional Average Treatment Effect (CATE) estimation have become increasingly accurate, they often remain black boxes, providing little insight into why treatments affect individuals differently. This paper introduces CATE-Concept Bottleneck Model (CATE-CBM), a novel framework that integrates concept-based learning with CATE estimation to bridge this interpretability gap. Our approach enforces a concept bottleneck that forces the model to express treatment effects through understandable concepts, enabling transparent reasoning about which concepts drive heterogeneous effects. Through experiments on a modified MNIST dataset, we demonstrate that CATE-CBM maintains competitive accuracy while providing meaningful concept-based explanations of treatment effect heterogeneity. The model successfully identifies how both the presence and absence of specific concepts influence treatment outcomes, offering clinicians and engineers both accurate effect estimates and interpretable rationales for personalized interventions. This work represents the first unification of concept bottleneck models with causal effect estimation, advancing the frontier of explainable artificial intelligence in causal inference. 10.18721/JCSTCS.18401 004.85 machine learning concept-based learning conditional average treatment effect interpretation neural network https://infocom.spbstu.ru/article/2025.87.1/

RAR RUS 20-29 0000-0001-7810-7973 Peter the Great St. Petersburg Polytechnic University Maleev Oleg

St. Petersburg, Russian Federation

Peter the Great St. Petersburg Polytechnic University Kovaleva Olga kovaleva.oa@edu.spbstu.ru

St. Petersburg, Russian Federation

A study of the applicability of the Kolmogorov–Arnold network architecture for time series forecasting The recently proposed Kolmogorov–Arnold Network (KAN) architecture emerges as a promising alternative to traditional neural networks based on the Multilayer Perceptron (MLP). By leveraging the Kolmogorov–Arnold representation theorem, KAN represents multidimensional functions as combinations of univariate functions, thereby offering potentially higher accuracy and model interpretability through its inherently simpler structure. This paper investigates the applicability of KAN to time series forecasting using the well-known hourly electricity consumption dataset as a benchmark. Meteorological observation data are selected as an additional testbed. A comparative analysis is conducted between KAN networks and traditional MLPs, as well as implementations of recurrent architectures based on KAN (TKAN variants) versus established designs such as Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU). Experimental results demonstrate the superiority of the KAN architecture over MLPs in temporal prediction tasks. The proposed recurrent architecture, TKAN1, achieves the highest coefficient of determination (R2 = 0.3483) among TKAN variants, with a Root Mean Squared Error (RMSE) of 0.1010 in energy demand forecasting. 10.18721/JCSTCS.18402 004.032.26 time series time series forecasting Kolmogorov–Arnold network multilayer per- ceptron recurrent neural network https://infocom.spbstu.ru/article/2025.87.2/

RAR RUS 30-43 0000-0002-1497-2893 Peter the Great St. Petersburg Polytechnic University Guo Chenxi chenxiguo.academic@gmail.com

St. Petersburg, Russian Federation

Potekhin Vyacheslav slava.potekhin@mail.ru Generative adversarial network for classification of mechanical fault diagnosis model The scarcity and imbalance of annotated fault data pose significant challenges to the reliability of intelligent industrial diagnostics. To address this issue, we propose an integrated fault diagnosis framework based on multi-domain feature fusion and generative adversarial networks (GANs). Unlike traditional approaches that treat generation and classification as independent stages, our model unifies these two processes. This method achieves diagnosis by transforming raw vibration signals into multi-domain representations (time domain, frequency domain, and time-frequency domain). The core innovation lies in the restructured generator architecture: a Transformer encoder captures global signal correlations, combined with an Efficient Channel Attention (ECA) mechanism for adaptive recalibration of feature weights, ensuring high-fidelity sample synthesis. Additionally, the model employs a dual-function discriminator that distinguishes genuine from synthetic samples while directly performing multi-class fault classification. Extensive experiments on CWRU and JNU benchmark datasets demonstrate that this approach surpasses existing state-of-the-art algorithms, achieving superior performance in Structural Similarity (SSIM), Peak Signal-to-Noise Ratio (PSNR), and diagnostic accuracy. This end-to-end solution effectively mitigates data scarcity challenges in industrial settings. 10.18721/JCSTCS.18403 004.8 fault diagnosis generative adversarial networks limited data supervised learning time-series analysis https://infocom.spbstu.ru/article/2025.87.3/

RAR RUS 44-52 Skiba Vladimir bauman@bmstu.ru 0000-0003-0644-1731 Petrenko Sergei petrenko.sa@talantiuspeh.ru Sirius University of Science and Technology Abakumov Evgeny abakumov.em@talantiuspeh.ru

“Sirius” Federal Territory, Krasnodar Krai, Russian Federation

Education system of engineers and scientific personnel in the sphere of quantum information technologies, quantum robust artificial intelligence and quantum resilience Currently, quantum technologies are the driving force behind technological progress (Industry 4.0/5.0/6.0 or Society 6.0). The development of quantum information technologies gives rise to new quantum threats to information security. The requirements of the federal project “Personnel for the Digital Economy or the Data Economy” for the key information technology – “Quantum Technologies” (Priority 1) – can be fulfilled only by developing an appropriate training system for engineering and scientific personnel in the fields of quantum information technologies, quantum robust artificial intelligence and quantum resilience. The creation of a vertically integrated education system, which includes not only basic higher education, but also the training of scientific personnel and upskilling for already employed professionals, will contribute to the development of innovative technologies in general. The paper presents the results of joint efforts of the Sirius University of Science and Technology together with the Quantum Consortium of Business and Universities to train personnel in the sphere of quantum information technologies for the development of such an education system. 10.18721/JCSTCS.18404 004.03 quantum computing quantum information technology information security educational program new quantum security threat quantum and post-quantum cryptography quantum robust artificial intelligence https://infocom.spbstu.ru/article/2025.87.4/

RAR RUS 53-66 Peter the Great St. Petersburg Polytechnic University Khuc Bang Thanh khucbang@mail.ru

St. Petersburg, Russian Federation

Gelgor Alexandr Low computational complexity technique based on a polyphase structure for modulation and demodulation of FBMC/OQAM-OTFS signals The paper proposes a low computational complexity technique based on a polyphase structure for modulation and demodulation of FBMC/OQAM-OTFS signals. This approach effectively reduces the overall computational complexity when compared to the frequency spreading approach (FS-FBMC/OQAM-OTFS), which in turn outperforms the direct modulation and demodulation approach for FBMC/OQAM-OTFS signals (Direct-FBMC/OQAM-OTFS). Simulation results explicitly demonstrate that, for an overlapping factor of K = 4, various PPN-FBMC/OQAM variants can indeed achieve a 2.5–4 times reduction in the computational complexity with an energy loss of no more than 1 dB compared to FS-FBMC/OQAM-OTFS. These obtained results are observed under standard multipath channel profiles such as EPA, EVA, and ETU in both moderately and highly dynamic scenarios. These findings suggest that the PPN-FBMC/OQAM-OTFS technique is a feasible and promising alternative to conventional OFDM in high-mobility wireless scenarios. 10.18721/JCSTCS.18405 621.391.8 OFDM OTFS FS-FBMC/OQAM-OTFS PPN-FBMC/OQAM-OTFS highly dynamic channel https://infocom.spbstu.ru/article/2025.87.5/

RAR RUS 67-75 Xi'an Jiaotong University Xu Luolan xuluolan123@gmail.com

Xi'an, China

Design of IoT device using beam-splitting prism display This project is based on an open-source hardware. Its functions have been redesigned and improved, with enhancements made to the hardware programming and power supply circuits. System-level secondary development has been carried out to implement an embedded system interface and a weather clock application. During development, modifying device functionality required close coordination between software and hardware and optimization of the enclosure structure. As resource demands increases, adjustments to the hardware power supply and software optimization become necessary to ensure reliable operation. Internet of things (IoT) device development necessitates a holistic approach. By working backward from the target specifications, hardware, software and enclosure design must be considered together. The results indicate that optimized power supply, low-coupling software operation and a thermally efficient enclosure significantly enhance the long-term stability and low-power operation of IoT devices. 10.18721/JCSTCS.18406 621.397 embedded systems development internet of things (IoT) devices beam-splitting prism display low-power optimization graphical user interface (GUI) https://infocom.spbstu.ru/article/2025.87.6/

RAR RUS 76-86 Peter the Great St. Petersburg Polytechnic University Gonzalez Mauricio F. fayula.gm@edu.spbstu.ru

St. Petersburg, Russian Federation

Peter the Great St.Petersburg Polytechnic University Antonov Alexander antonov@eda-lab.ftk.spbstu.ru

Polytechnicheskaya, 29, St.Petersburg, 195251, Russian Federation

Design and analysis of a reconfigurable hardware accelerator for solving a system of linear equations using Jacobi method This work presents the design and analysis of a reconfigurable hardware accelerator for solving a system of linear equations using Jacobi method, implemented on a reconfigurable device, as well as a comparative study of software and hardware implementations. Recent advancements in computing capabilities have been hindered by the so-called “walls”: memory, power consumption and clock frequency limitations imposed by current technology. Solutions to overcome these “walls” include reconfigurable computing and high-level synthesis. The system under development and analysis was described in the C++ language and implemented using a high-level synthesis method, which reduces design time and enables more efficient exploration of different hardware architectures. The comparative analysis showed a performance increment over the original implementation, with energy consumption comparable to that of a modern mid-class microprocessor. 10.18721/JCSTCS.18407 004.312.44 supercomputer reconfigurable hardware accelerator Jacobi method field-program-mable gate array high-level synthesis SystemVerilogHDL reconfigurable computing https://infocom.spbstu.ru/article/2025.87.7/

REV RUS 87-101 Peter the Great St. Petersburg Polytechnic University; Lomonosov Moscow State University Razuvaev Daniil Razuvaev_DD@mail.ru

St. Petersburg, Russian Federation; Moscow, Russian Federation

6603839750 Peter the Great St.Petersburg Polytechnic University Sergey M. Ustinov usm50@yandex.ru

Polytechnicheskaya, 29, St.Petersburg, 195251, Russia

Mitigating data growth in PoW blockchains: Storage reduction methods for scalability without compromising decentralization The exponential growth of data volume in Proof-of-Work (PoW) blockchains threatens their decentralization. The article provides a systematic analysis of data growth mitigation methods (sharding, block pruning, off-chain storage etc.), identifying their key flaws: compromised auditability, increased synchronization complexity or centralization. To address the issue, a novel Periodic Aggregation with Dual Hash Anchoring (PADHA) method is proposed. Its key innovation is the synergy of data pruning and the controlled use of chameleon hash functions. The method enables linear reduction of stored history by creating final state aggregators and subsequent secure “cleansing” of blocks from past epochs. PADHA preserves cryptographic chain integrity and PoW support without trusted third parties. The method is designed for Fact-Oriented Blockchains that store final data states (facts), making it promising for registries, IoT and other applications where current information, not its change history, is critical. 10.18721/JCSTCS.18408 004.9:004.94 distributed registries decentralized systems blockchain scalability decentralization block pruning https://infocom.spbstu.ru/article/2025.87.8/

RAR RUS 102-111 0009-0002-4444-1249 Saint-Petersburg State University of Aerospace Instrumentation Reshetnikova Nina

St. Petersburg, Russian Federation

V.A. Almazov National Medical Research Center” of the Ministry of Health of the Russian Federation Kuzmin Alexey kuzmin_as@almazovcentre.ru

St. Petersburg, Russian Federation

Saint-Petersburg State University of Aerospace Instrumentation Nikitin Alexander

St. Petersburg, Russian Federation

Saint-Petersburg State University of Aerospace Instrumentation Karamyshev Igor

St. Petersburg, Russian Federation

Application of VR-technologies for neurorehabilitation of patients with motor and cognitive function disorders The article explores the use of developed interactive 3D application with VR technology support, designed for the rehabilitation of patients after neurological disorders that cause motor and cognitive impairments. The neurorehabilitation hardware and software system is implemented on the Unity platform and includes an interactive 3D application and a Quest 3 VR headset. The VR headset consists of a virtual reality helmet and two controllers for the right and left hands, respectively, and supports highly accurate controller tracking using built-in cameras equipped with LiDAR technology. The causes of unilateral spatial neglect syndrome, or hemispatial neglect, and the possibility of its rehabilitation are considered. The hardware and software system developed by the authors enhances patient motivation through gamification and adaptive scenarios, allowing them to select and visualize various 3D scenes. When immersed in a realistic virtual environment, the patient acts instinctively. This helps develop hand motor skills, a sense of balance and spatial navigation abilities. Furthermore, the hardware and software system allows for the convenient storage of statistical data on the progress of prescribed procedures and can be used for further analysis during rehabilitation process. 10.18721/JCSTCS.18409 004.496, 616.8-085.851 cognitive functions motor functions hemispatial neglect rehabilitation virtual reality interactive 3D graphics game engine https://infocom.spbstu.ru/article/2025.87.9/

RAR RUS 112-122 Peter the Great St. Petersburg Polytechnic University Fershtadt Mikhail tral1930@mail.ru

St. Petersburg, Russian Federation

Shashikhin Vladimir Multi-criteria control of large-scale nonlinear dynamical systems without linearization, based on Lyapunov functions This paper proposes a numerical control method for large-scale nonlinear dynamical systems, focused on maintaining stability without using linearization. The approach under study is based on the principles of multi-criteria optimization, where the stability of the system is directly included in the vector of target criteria through Lyapunov functions. This allows us not only to minimize deviations from the target states and energy consumption for control, but also to guarantee the asymptotic stability of the system under arbitrary initial conditions. A mathematical formulation of the problem is presented, a discrete numerical control scheme is developed, and a scalarization strategy is proposed that provides an approximation to Pareto-optimal solutions. A series of numerical experiments implemented in Python has been conducted, confirming the effectiveness of the method using examples of both single- and multi-agent systems. The results demonstrate the stable behavior of the trajectories, a decrease in the Lyapunov function over time and correct operation even with strong nonlinearity of the model. 10.18721/JCSTCS.18410 519.8 nonlinear systems stability Lyapunov function multi-criteria optimization distributed control numerical simulation control without linearization https://infocom.spbstu.ru/article/2025.87.10/