Теоретичні та прикладні аспекти побудови програмних систем

Permanent URI for this community

https://ekmair.ukma.edu.ua/handle/123456789/38239

Browse

Browsing Теоретичні та прикладні аспекти побудови програмних систем by Title

Now showing 1 - 20 of 102
  • Thumbnail Image
    Item
    Accelerated Epidemic Simulation in Large-Scale Networks: Optimization of the Gillespie Algorithm and a Two-Layer Approach
    (Національний університет "Києво-Могилянська академія", 2025) Kuryliak, Yulian; Emmerich, Michael
    This study addresses the challenge of accelerating epidemic simulations in large-scale complex networks through algorithmic and structural optimization. Traditional Gillespie-based stochastic simulations accurately reproduce epidemic dynamics but become computationally prohibitive for networks exceeding tens of thousands of nodes. To overcome this limitation, we build upon such efficiency techniques as local rate updates and ordered event-selection structures, which reduce the computational complexity of each simulation step from O(n) to O(log(n)). Building on these principles, we propose a two-layer (micro–macro) modeling framework: the micro layer simulates intra-community dynamics, while the macro layer captures inter-community infections using hazard-integral rates derived from mobility data and the epidemic states of metanodes. This hierarchical approach enables scalable and parallelizable simulations that preserve stochastic accuracy while substantially reducing computational cost, allowing realistic modeling of epidemic spread across millions of agents and multiple cities.
  • Thumbnail Image
    Item
    Data Streaming Pipeline for the Quadcopter Flight Control Stack
    (Національний університет "Києво-Могилянська академія", 2025) Zavaliy, Taras; Shakhovska, Nataliia; Yatsyshyn, Volodymyr
    Modern UAV and UGV technologies are highly competitive dynamic field of applied research and development. Mainstream software stacks being used include PX4, ArduPilot, Betaflight, as well as ROS2 – a modular computing environment for processing sensor data streams, running perception and control algorithms, as well as software-in-the-loop simulations. In our research, ROS2 integration with Betaflight for on-board sensor data streaming is implemented and tested.
  • Thumbnail Image
    Item
    DecisioNet з пропорційним розподілом обчислювальних потужностей
    (Національний університет "Києво-Могилянська академія", 2025) Мокрий, Михайло; Швай, Надія
    У цьому дослідженні розглянуто нейронну мережу з бінарною деревоподібною структурою DecisioNet (DN) [1], яка належить до категорії нейронних дерев рішень [2] та представлено нову версію моделі з пропорційним розподілом обчислювальних ресурсів.
  • Thumbnail Image
    Item
    Design of an Intelligent Agent for Autonomous Microservice Optimization in Enterprise Systems
    (Національний університет "Києво-Могилянська академія", 2025) Vanin, Danylo
    This study introduces a conceptual framework for an intelligent agent designed to autonomously optimize microservice-based enterprise systems. Microservice architecture [3] serves as the basis for scalable and modular enterprise solutions, supporting agile DevOps practices and enabling independent deployment, updating, and management of services. While these features accelerate release cycles and improve maintainability, they also create persistent operational and architectural challenges [4, p. 633-634] that often necessitate manual intervention. These issues are especially significant in corporate settings, where requirements are dynamic and influenced by organizational, legal, and economic factors. In response, the software engineering field is increasingly exploring AI-driven approaches [1, p. 12]. Although machine learning has been utilized for anomaly detection [2], auto-scaling, and traffic routing, there remains a gap in the development of fully autonomous agents capable of intelligent architectural analysis and modification. The proposed agent addresses these challenges through a modular architecture that includes monitoring, decision-making, execution, and safety verification modules. The decision component employs reinforcement learning and formal quality models to balance objectives such as performance, cohesion, and latency, while maintaining adherence to system constraints. This design seeks to reduce architectural drift, improve service modularity, and support adaptive system behavior in real time. The agent is intended as a foundational element for intelligent DevOps and self-optimizing enterprise software systems.
  • Thumbnail Image
    Item
    Efficient Policy Learning via Knowledge Distillation for Robotic Manipulation
    (Національний університет "Києво-Могилянська академія", 2025) Severhin, Oleksandr; Kuzmenko, Dmytro; Shvai, Nadiya
    The work focuses on the computational intractability of large-scale Reinforcement Learning (RL) models for robotic manipulation. While world-like models like TD-MPC2 demonstrate high performance in various manipulative tasks, their immense parameter count (e.g., 317M) hinders training and deployment on resource-constrained hardware. This research investigates Knowledge Distillation (KD) with a loss function specifically described in [1] and [2] as a primary method for model compression. This involves training a lightweight "student" model to mimic the behavior of a large, pre-trained "teacher" model. Unlike in supervised learning, distilling knowledge in RL is uniquely complex; the objective is to transfer a dynamic, reward-driven policy, not a simple input-output function.
  • Thumbnail Image
    Item
    Enhanced image similarity detection: combining multi-layer outputs of CNN for precise results
    (Національний університет "Києво-Могилянська академія", 2024) Kubytskyi, Volodymyr; Panchenko, Taras
    The rapid growth of image data globally has amplified the demand for effective image similarity detection methods, particularly in tasks like image deduplication. This paper introduces a novel approach using enriched image embeddings derived from combining outputs of intermediate layers of pre-trained CNNs. The proposed method improves F1 scores across tasks such as near-duplicate detection, multi-angle view analysis, and schematical layout comparisons. Real-world applications in the real estate domain demonstrated fewer errors and enhanced performance, offering a promising direction for addressing complex image comparison challenges.
  • Thumbnail Image
    Item
    GPU-орієнтована бідіагоналізація в алгоритмі сингулярного розкладу з швидкою дефляцією та паралельним розбиттям
    (Національний університет "Києво-Могилянська академія", 2025) Сухарський, Сергій
    У роботі розглянуто GPU-орієнтований алгоритм SVD, що намагається усунути проблеми традиційних підходів завдяки масштабованій, опортуністичній дефляції на бідіагональній формі з негайним розбиттям на незалежні підзадачі.
  • Thumbnail Image
    Item
    Guided inverse problems
    (Національний університет "Києво-Могилянська академія", 2024) Ivaniuk, Andrii; Kravchuk, Oleg; Kriukova, Galyna
    The given work proposes a novel approach for solving inverse problems in machine learning leveraging Physics-Guided Neural Networks (PGNNs). This method incorporates domain knowledge through an additional inverse problem, leading to significant improvements in model performance and accuracy.
  • Thumbnail Image
    Item
    Information Education: Digital Knowledge and Online Language Learning
    (Національний університет "Києво-Могилянська академія", 2025) Didmanidze, Ibraim; Didmanidze, T.
    The presented paper outlines approaches to worldview researching language education in online environments and survey research that has been conducted to date. Is it worthwhile to use the Internet in Language Education? How does Language Education use change in on-line environments? What are the best ways to incorporate e-mail or the World Wide Web into the classroom? We can use several approaches to learn and research about such issues. One is to talk to fellow lecturers and professors. Another is to try practical things out in the classroom and see how they work. And another is to conduct and share scientific-theoretical research.
  • Thumbnail Image
    Item
    Information Practice Through Digital Ethics: Worldview Assessment of AI Reliability
    (Національний університет "Києво-Могилянська академія", 2025) Bagrationi, I.; Bagratishvili, A.
    The paper is devoted to the research and evaluation of ethical regularities of Artificial Intelligence as a certain digital existence in the context of worldview parameters. A critical review of the main provisions and principles surrounding the issue is given, finding and developing the progressive criteria in them and establishing a unique theoretical-worldview concept corresponding to modern requirements. The relationships fixed between the thinking of the fields of humanitarian and exact knowledge around the issue are shown in an original way; a kind of attempt is presented to understand and evaluate the system-conceptual model developed on the basis of moral criteria in the digital world of organized management in a new way.
  • Thumbnail Image
    Item
    Massively Parallel MorphoNAS Implementation
    (Національний університет "Києво-Могилянська академія", 2025) Medvid, Serhii
    Our GPU-accelerated MorphoNAS (Morphogenetic Neural Architecture Search) enhances the biologically motivated developmental model of the original MorphoNAS, from sequential developmental simulations to large scale parallel simulation. Like the original CPU-based MorphoNAS system, the GPU enhanced version models morphogenesis (neural development) through reaction-diffusion, progenitor differentiation and axon-guided wiring that grow recurrent controllers. Unlike the original CPU-based version, the GPU-based version has the ability to formulate growth and control in parallel execution of thousands of genomes. Through use of sparse, device-resident representations of developmental dynamics and recurrent rollouts, we produce behaviorally equivalent recurrent controllers with up to three orders of magnitude speedup. As a result, we have enabled the possibility of performing evolutionary search over populations of developmental programs, transforming MorphoNAS from a proof-of-concept model of artificial morphogenesis and adaptive architecture discovery to a scalable framework for those objectives.
  • Thumbnail Image
    Item
    ML-підхід до ідентифікації емоцій у звукових записах на базі MFCC
    (Національний університет "Києво-Могилянська академія", 2025) Тимошевський, Данило
    У сучасному світі, де взаємодія людини з комп’ютерними системами набуває все більш природного характеру, розпізнавання емоцій за голосом стає одним із ключових напрямів розвитку штучного інтелекту. Голос є одним із найвиразніших каналів передавання емоційного стану, тому аналіз аудіосигналів відкриває можливості для створення інтелектуальних систем, здатних розуміти не лише зміст висловлювань, а й їх емоційне забарвлення. Такі технології мають широкий спектр практичного застосування — від покращення роботи віртуальних асистентів, систем підтримки клієнтів і адаптивних освітніх платформ до психологічного моніторингу стану користувачів. Окремим напрямом, що набуває актуальності, є використання аналізу емоційного тону голосу для оцінювання ефективності ведення гри в настільних рольових іграх, де емоційна залученість та динаміка голосу відображають якість ігрового процесу та комунікації між учасниками.
  • Thumbnail Image
    Item
    Moore-Penrose Pseudoinverse Matrix
    (Національний університет "Києво-Могилянська академія", 2025) Kravchuk, Oleg; Kriukova, Galyna
    The Moore-Penrose pseudo-inverse is a foundational concept in modern numerical linear algebra, offering a principled approach to solving ill-posed and inconsistent systems arising in machine learning and other fields. This paper explores the pseudo-inverse from five distinct perspectives — axiomatic, variational, regularization, spectral, and algebraic graph theory — highlighting its theoretical depth and practical relevance across disciplines such as machine learning, signal processing, and network analysis.
  • Thumbnail Image
    Item
    Nowcasting як сучасний підхід до оцінювання ВВП України: порівняння з традиційними моделями прогнозування
    (Національний університет "Києво-Могилянська академія", 2025) Болотов, Єгор; Дрінь, Світлана
    Метою даного дослідження є порівняльний аналіз точності прогнозів ВВП України. До таких сучасних підходів до прогнозування на основі моделей змішаної частоти належать векторної авторегресії (MF-VAR) [1] та їх факторних розширень (MF-FAVAR) з традиційними моделями прогнозування. Особливий акцент робиться на їх здатності інтегрувати багаточастотні та нетрадиційні джерела інформації. Це дослідження є особливо актуальним, оскільки, хоча в українській науці вже існують дослідження змішаних частотних моделей на національному рівні (наприклад, U-MIDAS [2]), сучасні економічні підходи, такі як MF-FAVAR, ще не застосовувалися для регіональних прогнозів.
  • Thumbnail Image
    Item
    On algebraic properties of weakly locally compact abelian groups
    (Національний університет "Києво-Могилянська академія", 2024) Surmanidze, Onise
    By definition, a locally compact commutative group is a group that contains an open compact subgroup for which the factor group is discrete. In our earlier works, weakly linearly compact and locally weakly linearly compact groups are similarly defined. Their extension is represented by a weakly locally compact group.
  • Thumbnail Image
    Item
    On the way to universal executor
    (Національний університет "Києво-Могилянська академія", 2024) Svystunov, A.
    The work is dedicated to the universal executor architecture concept - an agentic system that can utilize both LLM-based inferences and computing by acquiring and memorizing programs.
  • Thumbnail Image
    Item
    Optimizing skin image segmentation with fourier and graph-based methods
    (Національний університет "Києво-Могилянська академія", 2024) Kinshakov, E.; Parfenenko, Yu.
    This paper introduces advanced methods for skin disease image segmentation using the Dermnet dataset, one of the largest resources in dermatology. Traditional approaches like Watershed and thresholding often fail due to the complex textures, color variations, and noise present in skin images. To address these challenges, novel techniques were proposed. First, the Fourier transform reduces high-frequency noise, preparing images for segmentation. Then, min-cut/max-flow graph algorithms minimize energy functions, enabling precise separation of pathological and healthy areas. Additionally, a piecewise smooth approximation improves boundary detection, refining segmentation results. Experiments demonstrated a 15% accuracy improvement over traditional methods. Processing time was also significantly reduced, enhancing the reliability and efficiency of automated diagnostic systems.
  • Thumbnail Image
    Item
    Philosophy of education: information system and applied ethical issues
    (Національний університет "Києво-Могилянська академія", 2024) Bagrationi, I.
    The present scientific report deals with the basic and fundamental theoretical aspects of axiological biases of some worldview issues and ethical problems in scope of education science through the transformation of digital technology system. The research idea of the report is that the process of digitalization of educational reality will be successful only with the support of the regulative practical ethical standards and values adequate to the formation of informational technologies. The report underlines that a computer, in principle, is not capable of converting concepts into meanings, information into knowledge. The report defines that the essence of "computer metaphor" is nothing more than a symbol; information processing by a computer is not a mechanism for generating knowledge from it by a person, and it is necessary to look for ethical patterns and mechanisms for understanding this process.
  • Thumbnail Image
    Item
    PINN Modeling of Interfacial Gravity-Capillary Waves
    (Національний університет "Києво-Могилянська академія", 2025) Avramenko, Olha; Sontikov, Maksym
    This paper presents an automated computational framework for modeling hydrodynamic processes using physics-informed neural networks (PINNs). The modular system integrates all stages of numerical experimentation — from data generation and model training to validation and accuracy evaluation — ensuring reproducibility, flexibility, and scalability. The framework was verified on the classical problem of interfacial gravity–capillary waves between two incompressible fluids, using the analytical solution as a benchmark for numerical assessment. Computational experiments showed that increasing the number of training points from 400 to 1000 improved accuracy and convergence, with the Extended configuration achieving 98.86% accuracy and a MAPE of 1.14%, while Adaptive_LR remained stable. The results confirm the reliability and efficiency of the proposed PINN-based framework for solving complex hydrodynamic problems governed by nonlinear partial differential equations.
  • Thumbnail Image
    Item
    Selection noise in genetic algorithms
    (Національний університет "Києво-Могилянська академія", 2024) Gulayeva, Nataliya; Borrego-Díaz, Joaquín; Sancho-Caparrini, F.
    In this study, authors concentrate on the selection noise characteristic of SSs. Recall that genetic drift, a well-known phenomenon in population genetics, is observed in GAs due to the stochastic nature of SSs. In a finite size population, a random selection among individuals of equal fitness leads to a disproportion between the expected and actual number of copies of an individual in the mating pool. We study selection noise of the most popular SSs used in generational GAs.