Harvester open publications of NAS Ukraine
Harvester open publications of NAS UkraineNot a member yet
2350 research outputs found
Sort by
Modelling videocard memory performance for LLM neural networks
No full textThe paper covers the analysis of performance characteristics of neural network-based algorithms class Generative pre-trained transformer, also known as Large Language Model, for contemporary videocards. The goal is to check the application limitations for this class of neural networks for mobile computing platforms. This network class is interesting for use as control system tool, but for the bigger text corpuses the network performance degrades and the number of used computer resources grows quickly, so we need to explore if this type of network, but based on a smaller text corpus, is a feasible tool for devices with comparatively low computing capability. The performance investigation was performed with the help of GPGPUSim simulator, which can be freely configured as a virtual videocard of any computing capability. As this neural network computations are based on the calculation of a sequence of matrix multiplications, and its performance is limited by the memory bandwidth, we analyze the behavior statistics of the different cache memory levels of the videocard processor and the cache interaction with the main memory. With the help of GPGPUSim we have gathered statistics for different Generative pre-trained transformer version 2 configurations, from small to xl configurations. The level 2 cache memory access statistics, level 2 cache memory access misses, number of accesses to main memory show that even for the middle-level network configurations the number of level 2 cache memory misses exceeds the 7-8% level. This number looks high and this evidence shows that the size of the cache memory is quite small for executing this neural network configuration, also there is the substantial traffic from cache memory to main memory. Although, the minimal so-called small configuration can be computed faster and with moderated resources, and so can be used further as a part of decision-making system for computing platforms with moderate performance and resources for the case of limited text corpuses. This opens good enough perspectives for using this type of neural networks for autonomous decision making.Prombles in programming 2024; 2-3: 37-4
Human face recognition system in video stream
Get PDFIn the work, an analysis of detection methods and faces in the video stream and their effectiveness in real time was carried out. Modern algorithms and pre-trained models have been found to be able to recognize faces with high accuracy, but their significant drawback is, in particular, vulnerability to attacks using fake faces. Therefore, the work also analyzed approaches to detecting living faces and the possibility of their implementation in the system. Using an object-oriented approach, a tool for face capture, receiving a video stream from various sources, detecting unknown and previously captured faces in a video stream, and recognizing live faces was designed and developed. The system has been adapted to work in real time using the GPU. The work improved the architecture of a convolutional neural network for recognizing living faces with the creation of a dataset from a combination of own footage and open datasets. Also, a user interface for the face recognition system was developed. The work improved identification procedures and simplified detection of persons on video for employees of the security department of enterprises by implementing liveness detection face recognition methods. As a result of the research, a system was designed, which is intended for detection, recognition and detection of living faces in a video stream. After analyzing the known successful software products, niches that need a new solution were identified. Based on them, functional and non-functional requirements were developed. The process of recognizing faces in the video stream has been modified by implementing our own Liveness Detection model.Prombles in programming 2024; 2-3: 296-30
Application of neural networks in OLAP-systems
Get PDFThe article highlights the main characteristics of OLAP systems that perform online analytical data processing. These systems, based on OLAP technology, are widely used both in government agencies and in private ones. The main characteristics, features and structure of OLAP systems are mentioned. The article emphasizes that OLAP is a data warehousing tool. OLAP allows analysts to explore and navigate a multidimensional structure of indicators called a data cube or OLAP cube. Indicators (measures) of OLAP cubes play an important role in the decision-making process. To solve some problems, these measures often need to be classified or clustered. Moreover, empty measures are common in OLAP cubes. Empty measures can present due to nonexisting facts in data warehouse or due to empty cells which are unfilled in by mistake. The presence of empty measures negatively impacts strategic decision making. Unfortunately, OLAP itself is poorly adapted for forecasting empty measures of data cubes. Over the years, researchers and analysts have tried to improve the decision-making process in OLAP systems and add forecasting and other options to OLAP applications. Today, in the era of Industry 4.0, with the availability of big data, there is a need to apply new technologies to solve such problems. These technologies include neural networks. The article examines the problem of integrating OLAP and a neural network. In this regard, the article provides information about neural networks: information about their properties, types, as well as their capabilities. The article shows the possibility and advantages of integrating OLAP and neural network. It mentions that in the case of big data, the integration of OLAP and neural networks is very effective for solving problems of classification, clustering and prediction of empty measures of OLAP cubes. An architectural and technological model for integrating OLAP and neural networks is presented. It is noted what types of neural networks can be used to solve the problems of classification, clustering and forecasting specified in the model.Prombles in programming 2024; 2-3: 367-37
Semantic technologies as a tool of information support for professionalization of andragogues
Get PDFThe article substantiates the expediency of using semantic technologies as a tool for professionalization of andragogues. Authors determine the basic preconditions that have an influence on the effectiveness of information support for the training and professional activity of andragogues. The clarity of the terminological apparatus in the field of learning and education of adults provides the basis for the further development of a formal model scheme of this subject area that interoperably fixes the various relations between concepts at the level of semantics. The online repository of relevant information resources that supports semantic structuring and unified integration of content provides convenient access to knowledge in the subject area. The availability of an appropriate set of advisory and retrieval services provides personalized information support for the training of andragogues using the content of this repository. The authors proposed the AndraMedia advisory system that is intended for process of informational support of the professionalization of andragogues. It ensures the integration with other intelligent advisory systems, open educational resources, online encyclopedias, ontologies, consulting chatbots and various external knowledge bases. AndraMedia is based on the Semantic Web standards of knowledge representation and processing. Ontological model of AndraMedia system formalizes the relations between the main objects and subjects of the domain of andragogues professional training. The use of pertinent ontologies provides an unambiguous interpretation of the semantics of information resources and their integration with semantically enhanced wiki technologies used to structure content and to search it at the semantic level. The software implementation of the repository is based on MediaWiki and its SemanticMediaWiki plugin. Prombles in programming 2024; 2-3: 441-44
Computer simulation of diffusion transport processes in multilayer nanofilms
Get PDFThe difficulties associated with studying diffusion in multilayer films require the progress of contemporar y modeling methods and software platforms to precisely represent phenomena, taking into account transitions between adjacent layers. In addition to the indispensable role of advanced modeling techniques and software in solving the problems of studying diffusion in multilayer films, it is extremely important to admit the key contribution of sophisticated computational approaches. In this paper, the authors attempt to merge intricate mathematical models with optimal software development methodologies to address the challenge of simulating diffusion transport processes in multilayer nanofilms computationally. Based on the experimental findings and employing the suggested model, identification was conducted utilizing the theory of state control for multicomponent systems. With the help of methods of optimal control of the state of multicomponent transport systems, the analytical solution of the model and the data of experimental observations, the distributions of diffusion coefficients for the considered components of nanofilms (samples of aluminum, molybdenum, silicon) were reproduced. Numerical simulation results were compared with experimental observations. The profiles obtained from the modeling closely match the corresponding experimental profiles, especially as the duration of multilayer formation converges to the final stages of completing the protective nanofilm multilayer formation. The maximum observed deviation does not exceed 2–3%, confirming the reliability of the mathematical model and demonstrating the practical value of the results provided. A software framework is developed for the automation of the specified calculations with the possibility of extension to other subject areas with similar tasks of identifying the key factors of the process and further numerical modeling of the time-space characteristics using the obtained results.Prombles in programming 2024; 2-3: 62-68
Agile requirement analysis approach using artificial intelligent technologies
No full textAn approach to requirements analysis using artificial intelligence technologies, taking into account the specifics of the AGILE methodology is proposed in this paper. The approach corresponds to the Model-Driven Methodology, in which the main artifacts of software development are software models represented by UML diagrams. The proposed approach corresponds to the key ideas of the AGILE manifesto, and is oriented towards the fact that AGILE has a priority to satisfy a customer when he changes requirements. Artificial intelligence technologies serve to prepare initial information for the “Text to Model Transformation” of the requirements specification into those types of UML diagrams (Use Case and Sequence), which are used for requirements analysis. The choice of the UML diagram visualization environment is substantiated.Problems in programming 2024; 2-3: 140-14
About one approach to automatic creation of formal queries to ontological knowledge bases
Get PDFThe article develops an approach that includes the analysis of short natural language messages in Ukrainian and the automatic generation of queries in SPARQL and Cypher based on them. The Apache Jena Fuseki server is used as a SPARQL query processing tool, and the Neo4J graph database is used as a data warehouse or ontological knowledge base. The latter is the most common open source database, highperformance and well-scalable, i.e., capable of working with large amounts of data. In addition, approaches to building formal queries based on natural language queries for Cypher are little known and require further development. The approach is based on the fact that a user's natural language query is subjected to a series of sequential checks. Their results determine the set of semantic types expressed in the phrase (natural language query) and the corresponding concepts that define them. The result of these checks is a set of four values – the codes of the check results, as well as the subjects and predicates, if present. This information is enough to select a set of basic templates for formal queries. Based on the results of such basic checks, the main basic templates for generating the final request are created. The proposed approach has a basic query template aimed at obtaining information of a certain type in a given form, as well as additional modifier templates that optionally construct query strings in the corresponding blocks of the main query by introducing additional conditions. The article describes the process of automatic generation of SPARQL queries to a contextual ontology using the example of a knowledge base of medical articles from peer-reviewed open access journals. The peculiarity of the approach is that the formal query is automatically built from blocks of templates (main and auxiliary), which are customizable in accordance with certain semantic categories present in the analyzed text and the entities that specify them.Prombles in programming 2024; 2-3: 326-33
Transforming governance: enabling scalable and adaptable decentralized networks on EVM-compatible blockchain
Get PDFThis article explores scalable and adaptable governance through decentralized networks, enabling collective decision-making and evolution without necessitating a complete system overhaul when original functionalities become obsolete. It delves into the utilization of interconnected Smart Contracts on the EVM-based blockchain to resolve foundational governance issues. However, the principal advantages are compromised if the system requires continual redeployment to adapt to environmental changes. We introduce a solution that combines a role-based access system and a modular system contracts architecture to enhance the system's scalability and adaptability. This approach allows for modifications and scaling at any time through community proposals and voting, catering to the specific needs of its members and eliminating the need for manual configuration by a centralized entity for new modules or functionalities. Members can propose configuration parameters for a new module, and with community majority approval, the system can adapt and scale, safeguarding the interests of its members, provided there is cooperation within the majority of the community.Prombles in programming 2024; 2-3: 351-358
Development of a methodology for the implementation of transactions in distributed systems with microservice architectura
No full textThe paper describes the analysis of the problems of using microservice architecture in distributed systems. Emphasis is placed on flexibility in the choice of technologies, scalability and organization of teams working on given microservices, technical and domain problems of transaction implementation in comparison with a monolithic system. The main focus is on transactions, as they ensure atomicity, consistency, isolation, and persistence across multiple services. In the process of analyzing modern approaches and solutions for working with transactions in distributed systems, it was found that one of the effective solutions is the use of the Transactional Outbox pattern. Its implementation in the form of Spring starter is presented. The latter is added to the system, configured and facilitates the use of transactions and the publication of events that are part of a transaction in a microservice architecture. The developed methodology for implementing distributed transactions based on message queues, using the above-mentioned starter, is described in detail. The basic configurations and settings of message queues for the correct operation of transactions in distributed systems are definedProblems in programming 2024; 1: 64-7
Проєкт оптимізації колекційно-експозиційної ділянки “Сирінгарій” Державного дендрологічного парку “Олександрія” НАН України
No full textThe paper presents the results of research of archival materials on the introduction of taxa of the genus Syringa in the “Оlexandria” State Dendrological Park of the National Academy of Science of Ukraine (Dendrological Park “Olexandria” NAS of Ukraine). Conventionally, two main stages of lilac introduction in the Dendrological Park “Olexandria” NAS of Ukraine were distinguished: the first stage – from the end of the 18th century until 1918 and the second stage – from 1946 until the present time.The laying of the monogarden “Syringarium” is considered to have appeared between 1948 and 1950. In 2021, the collection of “Syringarium” comprised two species, one artificial hybrid, and 37 cultivars and was represented by three age groups (i.e., I age group – 76 specimens of 19 cultivars; II age group – 90 specimens of 24 cultivars; III age group – 54 specimens of 15 cultivars).Considering over 70 years of lilac cultivation on one plot, it went through many difficulties, which resulted in changes in its diversity and a decrease in the quantitative composition of the collection. In particular, allelopathic soil fatigue caused significant weakening of plants and reduced their resistance to disease and pest damage. Therefore, there was a need to optimize the site of the “Syringarium” (i.e., partially replace the soil; re-root the plants on their own roots to extend their life; restore the varietal diversity by introducing the new cultivars, including those of Ukrainian selection, etc.). According to the project developed by the scientists of the Dendrological Park “Olexandria” NAS of Ukraine on optimization of the mono-garden “Syringarium”, it is planned to plant 91 specimens of 47 lilac. For the eco-educational purpose, it is also planned to create an area representing original lilacs species