Research at Department of Computer Science

The research topics cover a broad collection of soft computing and related areas and are focused on model development and implementation of methods for extracting knowledge from uncertain and vague data contained in large and possibly distributed databases and flexible quering. The topics include fuzzy logic (Vojtas), neural networks (Andrejkova), database analyssis and design (Vinar), computer networks (Jirasek), computer simulation methods and operations research in general (Kyselovic), design of information systems (Semanisin), problems of combinatorical structures (Geffert), or hypermedia (Vinar).

Fuzzy logic programming and flexible querying. The results obtained so far cover soundness and completeness of various fuzzy logic programming and resolution systems, especially with arbitrary finite approximations of connectives, fuzzy abduction for descision making coupled with linear programming for cheapest solution. Flexible querying is covered by results on various types of fuzzy unification including second-order logic which is used to extend this theory to meta-data and at the end to extend the flexibility of our query system to include the capability of providing answers to natural language queries and finding the appropriate access methods even through the user does not know about them. Work done on structural complexity and combinatorial optimization where many fundamental results have already been obtained may conceivably be useful in considering the efficiency of various query optimization algorithms.

Database design and analysis. This involves knowlwdge of database and CASE tools and programming languages and interface and communication standards. The language of logic is used to provide the possibility of quickly constructing trial versions of sophisticated query analysis, implementation and optimization systems to test the basic approaches before creating the final version. Models for detection of similar records in distributed databases are studied too.

Multimedia, Hypermedia. Hypermedia is commonly thought of only in the context of web pages, teaching software etc. Some preliminary results suggest the possibility of using them in a neglected area of making the output of software systems more user friendly by providing the option to produce output in hypermedia from linked to "resident textbooks".

Artificial neural networks. Predictions capability of neural networks (NN) was studied together with the topology of NN depending on the solved problem and from the research point of view very interesting problem of their topological adaptability (it means, the tpopogy of NN is adapted according to soslved problems). Incremental topology can be used for feed-forward NN, genetic algorithms for recurrent NN.

Recently there was worked on the following research projects that obtained grant support: