Artificial Intelligence
Scientists NDL of artificial intelligence and cognitive science (SHIK) and graduate students of the SAIT department under the leadership of Professor Oleksandr Mokin are engaged in the construction and analysis of generative adversarial networks (GAN) and diffuse models of artificial intelligence
- Transformation of the target class for the segmentation problem using U-GAN, VPI Bulletin, issue 1, p. 81–87, 2024.
- The problem of ensuring the consistency of the generation of diffusion models of deep learning (2023)
- Analysis of reprocessing methods of panoramic dental X-ray images for image segmentation problems, VPI Bulletin, issue 5, p. 41–49, November. 2023.
- Influence of hyperparameters of diffusion models on the quality of generation (2023)
- Methods of reprocessing panoramic dental X-ray images for deep learning problems (2023)
- Boundary artifact in generative adversarial networks (2022)
Scientists NDL SHIK, graduate students and students of the SAIT department under the leadership of Professor Vitaly Mokin are engaged in research in the field of large language models and improvement of chatbots using artificial intelligence technologies
- A new approach of cooperation of large language models and hint construction techniques to Chat-GPT for automatic solving of natural language problems in mathematics (2024)
- Review of techniques for increasing the efficiency of using large language models for solving applied problems (2023)
- Intelligent technology for detecting text deepfakes using large language models, VPI Bulletin, issue 1, p. 110–120, February. 2024
- Comparative analysis of capabilities of large language models Alpaca, Vicuna, Falcon based on transformer architecture (2023)
On the subject of analysis and processing of natural language text (Ukrainian and English) (NLP) using modern intellectual technologies, interesting applied problems are solved by scientists NDL SHIK, graduate students and students of the SAIT department under the leadership of Professor Vitaly Mokin. A priority long-term direction is the creation of the WISEST system (by analogy with the European WISE) on the state of water resources of a given river basin. The purpose of this system is, firstly, to automatically fill the system with data from web scraping and parsing of social networks, news sites, reports, books and other text information. Secondly, all information must be checked and mutually verified. Thirdly, the problems of analysis, classification, generalization and other classic problems of artificial intelligence should be solved based on this information.
- Automatic knowledge extraction from environmental reports with reference to time and spatial coordinates of water bodies , VPI Bulletin, issue 3, pp. 101-110, 2025
- Development of a spatially weighted criterion for georeferencing natural language texts to bodies of water . MN-2025, Jun. 2025.
- The method of augmentation of texts about the state of water bodies based on intellectual linking to multi-link geoinformation systems of named entities, VPI Bulletin, issue 3, p. 55–65, June. 2023.
- Construction of a dataset for training intelligent models of the web system with information about environmental problems and measures in the water bodies of the South Bug River basin WISEST-SBB (2023)
- Notation System for Comparing and Synthesis of Intelligent Key Phrase Extraction Methods for Ontological Models in Information Systems, ICT&SD 2022. Lecture Notes in Networks and Systems, vol 809. Springer, Cham.
- Intelligent methods of extracting key phrases from the text for building ontological models of information and search systems (2022)
- Information intelligent technology of automated processing of text natural language information (2021)
- Information intelligent technology of automated georeferencing of environmental text natural language information, Scientific works of VNTU [Electronic resource]. – 2020. – No. 4.
- The concept of intelligent NLP technology for georeferencing and classification of open text information about bodies of water (KUSS-2020)
Under the leadership of Professor Vitaly Mokin, research aimed at the development of methods and technologies for deciphering satellite and aerial photography using artificial intelligence and modern geo-information technologies has been ongoing for a long time.
- Intelligent technology for constructing building plans based on aerial photographs of their roofs, VPI Bulletin, issue 1, p. 101–109, February. 2024.
- Modern information technologies for recognizing the roofs of buildings on aerial photography (2022)
- Systematic analysis of the fragment sizes of aerial photographs of agricultural lands to search for anomalies in them using machine learning methods, VPI Bulletin. – Vinnytsia: VNTU, 2019. – No. 3. – P. 75-85.
- Information technology for finding unregistered places of entry of polluted water into the river using satellite and vector data of geoportals, Scientific works of DonNTU. Series: "Informatics, cybernetics and computing". – Pokrovsk, 2018. – No. 2 (27). - pp. 30-35.
Also, other tasks are solved within the limits student scientific circle professors of the Department of SAIT Vitaly Mokina and Oleksandr Mokina "Applied information technologies, system analysis and artificial intelligence".