Search academic texts

Information × Registration Number 0224U031471, 0121U110109 , R & D reports Title To develop fuzzy and probabilistic machine learning methods based on high-performance computing popup.stage_title Head Yershov Serhii V., Доктор фізико-математичних наук Registration Date 30-04-2024 Organization V.M. Glushkov Institute of Cybernetics of National Academy of Sciences of Ukraine popup.description2 The object of the study is the methods of machine learning of fuzzy models of various types that describe cause-and-effect relationships between objects in a fuzzy information space and agent-oriented software systems for intellectual analysis of linguistically and stochastically uncertain expert and experimental information. The purpose of the work is to develop fuzzy and probabilistic machine learning methods for building complex models and algorithms for predicting the state of complex systems based on network models of connections between objects in a fuzzy information space and high-performance computing. Its research methods are methods of fuzzy mathematics, fuzzy logic, machine learning, agent-oriented programming and designing software systems and their components. Algorithms for machine learning of higher-order fuzzy systems have been developed, providing simplification of fuzzy rules and increasing the speed of fuzzy inference in a high-performance computing environment. Fuzzy agent-oriented machine learning methods for loosely structured subject areas have been developed. The method of building intelligent multi-agent systems based on fuzzy object-oriented dynamic networks and the architecture of software tools for these types of networks have been developed. The basic algorithms of fuzzy object-oriented dynamic networks have been developed, which allow for machine learning to calculate the degree of similarity between new knowledge and previously obtained knowledge in terms of fuzzy classes of objects. A method for identifying a priori estimates in Bayesian belief networks based on tiered structuring of the graph has been developed, which simplifies the parallel organization of calculations in machine learning based on fuzzy information. An experimental version of the software for machine learning of multi-level hierarchical systems of fuzzy logical inference based on high-performance computing systems has been built. Product Description popup.authors Veriovka Olha V. Hlushkova Vira V. Terletskyi Dmytro O. popup.nrat_date 2024-04-30 Close