The International Conference on Security, Fault Tolerance, Intelligence

NASCA METHOD: A COMBINED CNN-CVAE METHOD FOR MULTI-CLASS RESPIRATORY DISEASE DETECTION FROM AUSCULTATION DATA

2025-04-29T18:06:05+00:00

This paper presents the NASCA method, which combines convolutional neural networks (CNN) and convolutional variational autoencoders (CVAE) to enhance the classification of lung auscultation sounds under limited and imbalanced data conditions. The method integrates data preprocessing, synthetic augmentation, and robust deep learning classification with strict control against data leakage. Experiments conducted on the ICBHI dataset achieved an F1-score of 98.45% in multi-class classification. The results confirm the potential of NASCA for improving early respiratory disease detection.

IMPROVING REINFORCEMENT LEARNING FOR QUADRUPED ROBOT LOCOMOTION

2025-09-07T11:26:54+00:00

This article addresses the issue of machine learning of a quadruped robot for flexible locomotion, in slower learning environments with limited computational resources. To create a solution, reinforcement learning and a number of techniques are used to improve the result and speed of machine learning.

METHODOLODGY FOR STATISTICAL LEARNING OF NONLINEAR MODEL PARAMETERS WITH THE INTEGRAL FORM OF OLS IN THE SCHEME OF DIFFERENTIAL-NONTAYLOR TRANSFORMATIONS

2025-09-07T11:31:37+00:00

The article presents the methodology for statistical learning of nonlinear model parameters, its principle and efficiency of application. The results of modeling are shown with the statistical characteristics of the obtained expectation compared with the results of the application of the statistical learning methodology using the differential spectra balance.

SENTIMENT ANALYSIS OF NON-ENGLISH TEXT BASED ON NLP TECHNOLOGIES

2025-09-07T11:32:08+00:00

The paper addresses sentiment analysis techniques for texts in languages other than English using natural language processing (NLP) technologies, using Ukrainian texts as an example. A system for classifying the sentiment of Ukrainian news headlines using deep learning methods is presented and two LSTM-based neural network models are developed: one trained on a small manually-labeled Ukrainian text corpus, and another trained on a larger Twitter-based English corpus that was machine-translated into Ukrainian. A comparative evaluation of these models’ performance is conducted, and observations are made about language-specific challenges in sentiment expression for Ukrainian. The advantages and limitations of each approach are discussed and directions for further research in multilingual sentiment analysis are outlined.

AI DRIVEN LOG ANALYSIS IN COMPLEX SYSTEMS

2025-09-07T13:10:20+00:00

The article presents an overview of existing approaches to the application of various AI models and methods for log analysis in complex systems. A comparative analysis of the existing methods is conducted in terms of approaches and functional capabilities.

CONSTRUCTION OF A SET OF DISJOINT PATHS WITH QOS PARAMETER CONSIDERATIONS

2025-09-07T11:41:48+00:00

The paper contains research of construction of a set of disjoint paths. Construction of a set of disjoint paths provides various benefits and allows to perform load balancing. But most of the methods which were reviewed did not include paths satisfying Quality of Service requirements. Others include but ignore type of traffic. This paper presents a solution, which is designed to address this problem by constructing a set of disjoint paths that satisfy QoS metrics limitations considering type of traffic.

METHOD OF AUTOMATING WEB APPLICATION DEVELOPMENT

2025-09-07T11:45:43+00:00

The paper is devoted to the study of web application automation and the means to achieve it. Automation of development reduces costs, simplifies maintenance and updates, and allows focusing on business logic. However, selecting the optimal tool among the variety of available solutions remains a challenging task. The article proposes an automation approach based on the integration of modern frameworks and tools, which enhances development efficiency and improves integration with other systems.

INCREASING THE EFFICIENCY OF DATA PROCESSING USING QUANTUM COMPUTATIONS

2025-09-07T12:49:05+00:00

This article explores how emerging quantum computing techniques can be integrated into classical ETL pipelines to improve the efficiency of key data extraction, transformation, and validation operations through hybrid quantum-classical architectures.

QOS MANAGEMENT IN EDGE NETWORKS USING SDN CONTROLLERS

2025-09-07T12:55:27+00:00

This paper discusses approaches to optimizing traffic and resources at the edge of the network using software-defined networking (SDN) that provide flexible centralized management. The focus is on increasing data transmission efficiency, reducing latency, and ensuring consistent quality of service in distributed edge infrastructures.

IMPROVING QUALITY OF SERVICE IN SDN NETWORKS BASED ON DPI CLASSIFICATION AND CBQ+WRED

2025-09-07T13:04:26+00:00

This paper explores an approach to Quality of Service (QoS) management in SDN networks by implementing a combined CBQ+WRED mechanism with prior DPI-based traffic classification. It is demonstrated that the integration of priority-based packet handling and probabilistic early drop under software-defined routing control leads to reduced packet loss, delay, and jitter for critical traffic classes. Experimental results obtained through discrete-event simulation confirm the effectiveness of the proposed solution compared to traditional FIFO queuing.

PROGRAM PIPELINING OF GPU USING C-SLOW RETIMING

2025-09-07T11:17:34+00:00

The software pipelining methods used in various computer architectures are discussed. In particular, thanks to software pipelining, it is possible to significantly increase the performance of graphics processing units. The proposed software pipelining method is based on the creation and optimisation of a synchronous data flow graph through its folding using C-slow retiming, with its subsequent description in a programming language.

METHOD FOR ACCELERATING THE COMPUTATIONAL IMPLEMENTATION OF MULTIPLICATIVE OPERATIONS IN MODULAR ARITHMETIC

2025-09-07T11:51:49+00:00

This paper explores methods for improving the efficiency of implementing modular arithmetic operations, which are fundamental to public-key cryptographic systems. The research is motivated by the growing importance of information security and the increasing use of low-bit microcontrollers in real-time applications. The paper focuses on accelerating modular multiplication and squaring operations through optimized algorithms tailored for processors with limited bit capacity. The proposed methods leverage Montgomery reduction and segmented multiplication with dynamic operand adjustment, aiming to significantly improve computational speed while maintaining cryptographic robustness.

METHOD FOR HOMOMORPHIC ENCRYPTION OF IMAGES DURING REMOTE CLOUD PROCESSING

2025-09-07T11:57:38+00:00

The master's thesis examines the theoretical foundation and development of an efficient method for accelerating the Discrete Fourier Transform (DFT) using cloud computing for image processing. The key feature of the proposed approach is the use of bitwise modifiers to encrypt individual bits of the binary representation of the signal. This technique enables lightweight and flexible encryption on terminal devices, significantly complicating unauthorized reconstruction of the signal during processing on a remote computing system. A thorough theoretical analysis and experimental modeling were conducted, demonstrating that the developed method achieves a 1–2 orders of magnitude increase in processing speed compared to traditional schemes, while maintaining a high level of data confidentiality. The proposed solution is suitable for secure and efficient image processing in cloud environments, particularly in scenarios with limited computational resources on the terminal side.

METHOD OF USING CLOUD SERVICES IN COMBINATION WITH CI/CD FOR THE SMART CITY SYSTEM

2025-09-07T12:02:05+00:00

The article explores the feasibility of implementing cloud services in combination with continuous integration and delivery (CI/CD) technologies for the development of a flexible and scalable Smart City system. It analyzes the architectural features of cloud platforms, particularly Azure, and their role in deploying containerized applications. The advantages of using CI/CD approaches for automating the development lifecycle of Smart City solutions are presented. Practical aspects of integrating IoT devices, processing data in the cloud, and implementing monitoring and logging systems are discussed, with the goal of improving the reliability and manageability of the infrastructure.

METHOD FOR INCREASING THE EFFICIENCY OF SOLVING TRIDIAGONAL SLE IN A REDUNDANT NUMBER SYSTEM ON FPGA

2025-09-07T12:07:03+00:00

The article considers a new approach to the hardware implementation of the run method for tridiagonal SLEs using a non-autonomous computational mode and a redundant number system. A computing module architecture and a modified algorithm have been developed, providing acceleration of more than three times compared to the classical approach. The effectiveness of the proposed method for systems of different dimensions has been experimentally confirmed.

METHODOLOGY OF USER INTERFACE DEVELOPMENT BASED ON NFML MARKUP LANGUAGE

2025-09-07T12:14:48+00:00

The article presents a methodology for developing a user interface using the NFML markup language. It reveals the working principle, which involves transforming NFML code into the source code of the target platform, which can later be updated according to user needs. A comparative analysis with existing similar development methods is conducted, and the method's functionality is tested for HTML and Java platforms.

MODIFICATION OF END-TO-END VIDEO GAME TESTING METHODS

2025-09-07T12:17:58+00:00

The paper is devoted to the research of video game test automation and means to achieve it. Test automation helps speed up the development process and create video games more efficiently. But rates of test automation in the industry are low, because of the current methods’ disadvantages. The article overviews End-To-End layer of testing. The solution provides a test approach that is easier to maintain during active development, and more resilient to changes.

MULTICRITERIA SCORING METHOD FOR TRAVEL ROUTES BASED ON GRAPH MODELS

2025-09-07T12:23:07+00:00

The work involves the development and investigation of a method for multicriteria scoring of movement routes on graph-based models. The relevance of the topic arises from the urgent need for fast and reliable route planning for cargo and passenger transportation in wartime conditions, where the priorities are operational speed, safety, and resource efficiency. Modern navigation systems often fail to simultaneously account for multiple critical factors such as route risks, actual road surface condition, congestion, or travel restrictions. Applying multicriteria analysis based on graph models allows for a comprehensive evaluation of alternative paths and supports decision-making in dynamic and complex transport environments.

SIMPLIFYING THE DEPLOYMENT OF IOT SYSTEMS THROUGH MODULAR DESIGN

2025-09-07T12:39:39+00:00

The paper explores the potential of simplifying the deployment of IoT systems by establishing a single open standard for modular architecture. It examines how the lack of unified standards in IoT leads to fragmented solutions and deployment challenges. The paper reviews existing attempts at IoT modularity as limited forms of standardization, highlighting the need for a comprehensive open standard

MISSION CONTROL METHOD FOR THE SWARM OF DRONES BASED ON PETRI NETS

2025-09-07T13:00:47+00:00

The article describes and analyzes the process of development of a mission control method for the UAV swarm. The method is based on Petri nets.