OPOleksandr Poplavskyi

Oleksandr Poplavskyi

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Oleksandr Poplavskyi has a total of 25 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Multi-objective optimization of serverless function configuration using genetic algorithms

    by Oleksandr Poplavskyi, Dmytro Romanenko

    Serverless computing, specifically the Function-as-a-Service (FaaS) model, provides a scalable foundation for deploying modern applications. However, selecting the optimal memory configuration for AWS Lambda functions remains a challenging task due to the coupling of allocated memory size and CPU share, which creates a non-trivial trade-off between execution duration and monetary cost; classic tuning tools, such as AWS Lambda Power Tuning, ignore the multi-criteria nature of the problem and require an exhaustive search of the configuration space. The aim of the study is to develop an approach for the automated search of Pareto-optimal configurations of FaaS functions without an exhaustive search of the configuration space. The paper proposes an approach to multiobjective optimization based on the non-dominated sorting genetic algorithm NSGA-II with adaptive mutation, memoization of previously measured configurations, and elitist selection, which automatically determines the set of Pareto-optimal configurations. To overcome the performance variation inherent in cloud environments, a robust evaluation strategy has been implemented involving JIT compiler warmup, dropping the first invocation, and median filtering of ten consecutive measurements. Experimentally, on three workloads classes – computational, mixed, and memory-bandwidth-limited – it is demonstrated that the proposed evolutionary approach efficiently converges to a diverse Pareto front, exploring only 23–30% of the search space out of 182 configurations. A "knee point" is consistently identified near the 1769 MB threshold, which corresponds to the allocation of one virtual CPU; cost savings relative to the time-optimal configuration amount to 41–43% for computational workloads, 15–26% for mixed, and 22–39% for memory-bandwidth-limited ones. The scientific novelty lies in the comprehensive combination of NSGA-II with a robust evaluation strategy tailored for cloud environments with unstable performance. The practical significance of the approach is the ability to make informed decisions between high-performance and cost-effective FaaS configurations without exhaustive manual testing, thereby reducing operational expenses for industrial deployments.

  • Open Access

    Highly efficient approaches to processing complex visual data in decision support systems

    by Oleksandr Poplavskyi, Sergii Pavlov, Oksana Bezsmernta, Iryna Gerasymova, Bakhyt Yeraliyeva

    Modern decision support systems (DSS) increasingly must analyze large-scale, complex, and heterogeneous data streams in real time. High-performance, AI-driven processing methods – particularly deep neural networks – offer effective solutions. This study examines the integration of contemporary architectures – YOLOv8, ResNet-50, EfficientNet-B3, and the Vision Transformer (ViT) – to enhance DSS capabilities. The models are benchmarked on a representative image-classification task using the COCO dataset for training and evaluation. Empirical results indicate that the transformer-based model (ViT) attains the highest accuracy, whereas the one-stage architecture (YOLOv8) achieves the fastest inference. EfficientNet-B3 and ResNet-50 exhibit intermediate trade-offs between accuracy and speed. Deployment considerations across DSS scenarios are outlined: YOLOv8 is appropriate for real-time, resource-constrained environments; ResNet-50 provides balanced performance; EfficientNet-B3 offers strong accuracy with moderate computational demand; and ViT delivers the best accuracy when ample data and computational resources are available. The findings are discussed in the context of DSS workflows, illustrating how the model outputs can directly inform and improve decision-making processes.

  • Multimodal detection of urban improvement indicators in public visual-text content using deep learning

    by Oleksandr Poplavskyi, Yuliia Riabchun

    The subject matter of the article is the automatic identification of user generated content about urban infrastructure improvement and construction within very large streams of social media and group messages, with a focus on filtering actionable posts from overwhelming background noise so that experts can see what needs restoration or investment first. The goal is to design, implement, and validate a multimodal deep learning system that receives any image and its accompanying text and decides whether the pair is relevant to city improvement topics such as roads, parks, buildings, lighting, cleanliness, accessibility, or playgrounds, and to provide a practical content filter that reduces manual triage for municipal teams. The tasks to be addressed are to base training on a single large scale public corpus rather than collecting new data, to define a clear relevant versus irrelevant decision target, to construct a multimodal neural architecture that learns from both visual content and written description, and to evaluate the approach against single modality baselines under the same conditions. The methods used rely on the Wikipedia based Image Text dataset known as WIT, which contains more than thirty-seven million image text examples with entity rich captions from many languages and is available for download on the Hugging Face platform, and on a fusion of a convolutional or vision transformer backbone for images with a transformer-based encoder for text that together form a unified classifier trained with standard supervised learning. Conclusions. Experiments show that the proposed approach accurately pinpoints posts related to urban improvement and that the multimodal design clearly surpasses the single modality baselines in the same conditions, which confirms that combining image evidence with textual context is advantageous for this filtering task and supports practical deployment as an automated screening stage. Scientific novelty lies in applying multimodal deep learning to the specific problem of real time content filtering in the urban planning domain, in framing the target as detection of relevant civic information within public image text communications using a single widely available corpus for training, and in demonstrating that a simple but principled fusion of computer vision and natural language processing can distill actionable information from very large volumes of online messages without manual collection of new training data.

  • Open Access

    Methodological foundations of somatoscopic research in automated decision support systems

    by Олександр Поплавський, Oleksii Shkuratov, Serhii Kolisnyk, Анна Поплавська, Dmytro Yovenko, Сергій Колісник, Дмитро Йовенко

    Automated Decision Support Systems (ADSS) are widely applied in various fields of science and technology, particularly in medicine, where their role in diagnostic, prognostic, and therapeutic processes is undeniable. The use of computational technologies in medical practice has become a necessary stage in the development of the field; however, the increasing volume of data that requires processing, along with heightened demands for accuracy, speed, and reliability of recommendations, significantly complicates decision-making processes. The growing amount of data, combined with the need to minimize the risks of erroneous decisions, drives the scientific community to seek innovative information technologies that can ensure a high level of accuracy in computational operations while minimizing the time required for their execution. Somatoscopic measurements, which are a crucial element in assessing the condition of the musculoskeletal system and overall physical posture of a patient, require precise and systematic image analysis that allows the identification of critically important anatomical markers. This process involves the accurate determination of key points on images of anatomical structures, which serves as the foundation for making precise predictions about the patient's condition and subsequent treatment planning. In the context of continuously increasing volumes of medical data and stricter requirements for the quality of decisions, the implementation of innovative technologies, particularly deep learning, becomes increasingly relevant as it enhances the processes of analysis and diagnosis. The aim of this study is to conduct a comprehensive analysis of the methodological aspects of applying deep learning in the context of automating the diagnosis of the musculoskeletal system and posture assessment. The study emphasizes the use of specialized neural network architectures that enable the identification of key anatomical points on images, thereby facilitating a deeper and more accurate analysis of anatomical structures. This approach significantly improves the efficiency of diagnostic processes by minimizing the likelihood of errors in decision-making and optimizing the performance of medical systems. The results of the conducted research demonstrate the significant potential of deep learning algorithms in medical systems for the automated analysis of images, allowing for substantial improvements in the accuracy and speed of decision-making. Such automation contributes to reducing the risk of subjective errors associated with the human factor, which is particularly important in complex clinical cases. Therefore, the further development of research in this direction is of paramount importance for the medical field, as it opens new opportunities for solving complex diagnostic tasks at an innovative technological level. The integration of deep learning technologies into the processes of extracting somatoscopic data not only enhances diagnostic efficiency but also creates prerequisites for the development of new decision support systems that optimize medical practice.

  • Open Access

    Modern methods of risk management in Ukrainian energy markets

    by Олександр Поплавський

    In the context of rapid technological development and increasing instability on global markets, the issue of developing effective risk management methods on Ukrainian energy markets becomes particularly important. Modern risk management methods include the use of machine learning algorithms to predict and minimize potential threats. This study focuses on the implementation of innovative machine learning and artificial intelligence technologies to anticipate risks in energy markets and respond timely to potential stresses. This paper explores the development of intelligent risk management systems that apply deep learning neural network technologies for predicting market indicators. These systems provide detailed analysis of large data sets, opening new perspectives for the accurate identification of risk factors. The study involved analysis of historical data and assessment of current market trends, which allowed the development of a comprehensive approach to identifying price volatility in risky market segments. Results showed that the application of machine learning algorithms and artificial intelligence methods can significantly enhance the accuracy of risk prediction and optimize decision-making processes. It was found that intelligent systems with artificial intelligence elements enhance the understanding of market mechanisms and enable government officials, regulatory bodies, and management of energy companies to make more informed and effective management decisions. The findings indicate a great potential for the application of artificial intelligence and machine learning in risk management on energy markets. Utilizing these technologies can significantly reduce risks and enhance stability in resource management under unpredictability, which is key to ensuring long-term success in the field.


  • Test Design Methodology for Software Verification

    by Oleksii Lopuha, Svitlana Tsiutsiura, Oleksandr Poplavskyi, Oleksii Lysytsin, Olena Bondar, Pavlo Kruk

    This paper provides outlines the main points covered in a paper that provides a brief summary of software testing research. While the information is concise and informative, there is certainly room for expansion and further elaboration on the topic. To begin with, software testing is a critical component of the software development process. It involves the evaluation of software to ensure that it meets specific requirements, functions correctly, and is free of defects. The role of testing is to identify potential issues or errors in the software before it is released to end-users. There are various methods and models used in engineering testing of data processing software. Each of these methods focuses on a specific aspect of the software and aims to identify potential issues at different stages of development. In addition to testing methods, it is essential to have a structured testing process in place. Moreover, it is necessary to evaluate the testing process to determine its effectiveness and identify areas that may need improvement. Overall, software testing is a complex process that requires careful planning, execution, and evaluation. It is essential to use appropriate testing methods and models to ensure that software meets its intended purpose and functions correctly. By implementing a structured testing process and evaluating its effectiveness, software development teams can improve the quality and reliability of their products, leading to greater user satisfaction and increased business success.


  • Development of segment classification criteria based on the features of compression algorithms

    by Kvyetnyy, O.Yu. Sofina, R. Maslii, A. Olesenko, O.A. Poplavskyy, A. Smolarz, A. Sagymbekova

    This article suggests classification criteria of image segments which constitute the basis of a multicriteria model of segment classification by compression algorithms. This model allows development of a compression algorithm on the basis of existing popular algorithms which will be able to process several most prevailing image classes.

  • Open Access

    Method for automatic assessing of the gradient filter threshold for fast processing of dynamic scenes objects

    by Leonid Timchenko, Oleksandr Poplavskyi, Natalia Kokryatskaya, Anna Poplavska

    Method for automatic assessing of the gradient filter threshold for fast processing of dynamic scenes objects