Research Interests
The research interests and activities as falling under the umbrella of foundations of Computational Intelligence (CI) involve key information processing technologies such as fuzzy sets, Artificial Intelligence, neurocomputations, genetic algorithms, and evolutionary techniques including genetic programming. They are aimed at the development of hybrid intelligent systems that exhibit different levels of learning, seemingly combine explicit knowledge representation with the significant learning faculties and are capable of coping with uncertainty. While the above constitute a conceptual framework within which more specific methodological topics are pursued, there are several well defined and highly coherent research streams:

  • Software Engineering. The main thrust here is in the development of quantitative software metrics (measures), software quality models, and software reusability. The latter essential issue is considered in the setting of unsupervised or partially supervised learning that could help accommodate more specific design and retrieval preferences. Another pursuit is in the realm of the application of Computational Intelligence to software design and analysis. Here constructed are neurofuzzy models of software quality and cost estimation

  • System modelling and knowledge discovery. The main interest is in knowledge-based modelling that exploits fuzzy sets with a primary objective to develop user-friendly (or being more specific user - customizable) models. This, in turn, calls for the comprehensive studies on representing heterogeneous data of variable granularity, relation-based linguistic models, distributed and hierarchical models and patterns in data. Fuzzy modelling becomes central to most of the current applications of fuzzy sets including fuzzy controllers. In this specific research area the main focal points of research activities embrace

    • analysis of the controllers with respect to their robustness and fault tolerance,
    • model -based design of the controllers,
    • hybrid and hierarchical control or decision-making structures.
    • fast dedicated hardware architectures of fuzzy neural networks in
    • applications to ATM networks

  • Reconfigurable and evolvable architectures. With the increasing complexity of systems, an interesting and promising pursuit is to design systems through evolution and self-organization. Here the main thrust is in the exploitation of evolutionary optimization and their direct usage in evolvable hardware environment provided by FPGAs and FPGAAs.

  • Pattern recognition. Here a particular emphasis is focused on unsupervised (clustering) and partially supervised pattern classification. These studies concentrate on various generalizations of standard objective-function based clustering methods (such as e.g., Fuzzy Isodata) and Kohonen self-organizing maps. The generalizations are primarily concentrated on different mechanisms of partial supervision ( including various ways in which information about class membership is provided) and diverse types of patterns taken into consideration (numerical, set-oriented, or linguistic).