The main research trend at the Department is Soft Computing for IT.
Soft computing is not a single integrated field, it rather includes several disciplines that can be developed individually. However, grouping the disciplines under one paradigm makes it possible to seek and find interdependencies and employ the methods of one discipline in another discipline. Soft computing is characterised by contemporary development of the interaction between the computing theory and its applications.
The research activities are carried out in 4 areas:
Intelligent systems and algorithms
The project is focused on the research of artificial intelligence, mainly the methods of soft computing and the adaptation of individual elements of multi-agent systems representing intelligent systems. The methods of the analysis of multivariate data are linked with the results of our research team members. A very important branch related to artificial intelligence and data processing is pattern recognition. Modern recognition systems not only remember sample situations but based on them, they can make a general conclusion. Doing this, they attempt to develop a suitable model for the classification of objects into classes. In this manner, we are able to ensure and classify selected patterns that can define more than just the behaviour of the system itself. Hybrid approaches based on soft computing methods (neuron networks, evolutionary techniques, decision making under uncertainty, etc.) allow establishing interdisciplinary adaptive methods using the benefits of individual approaches. Established approaches to adaptive control of intelligent systems are verified on selected systems, e.g. autonomous robotic systems, intelligent houses, etc.
Application of fuzzy approaches for the analysis, description, prediction and control of systems
This research trend focuses on processing vague (uncertain) information that originates in the course of the analysis, prediction and control of systems. The aim is to develop fuzzy modeling tools in selected problem domains. The most widely used tool is Linguistic Fuzzy-Logic Controller, which has been developed in the Institute for Research and Applications of Fuzzy Modeling, the University of Ostrava.
The primary aim is further development of the Linguistic Fuzzy-Logic Controller tool on the basis of its applications. What is found to be most significant is the transition to using standard XML document when describing a set of fuzzy rules, and following system extension by a web-based communication interface. This will make it possible to employ a web application for editing the linguistic description for LFLC and the following communication with the LFLC tool from web client, which requires entering input values, starting LFLC with the linguistic description and the visualisation of outputs from the LFLC tool.
This will subsequently allow the application of this tool along with others such as F-transformation for the analysis of system experimental data, and for the reduction of noise and other negative effects. Vague information is then employed in the description of the system using the fuzzy-expert system and its application for the prediction of system behaviour and also for the system control in order to reach optimal results. Following the expert knowledge of the system behaviour, it is possible to reach (mainly when controlling a program) important suppression of the effect of transport delay etc. Application domains of developed fuzzy tools will mainly comprise the identification of the techniques of the use of search methods detecting cyber-attacks on networks in data captured by means of experimental honeypots operated by the team members, with the objective to reach the highest effectiveness of the detection of new attacks and respond in a timely way to them.
Another aim will be the development of an adaptive e-learning system controlled by a hierarchical system of several fuzzy expert systems.
Selected problem domains and solved areas are:
Linguistic Fuzzy-Logic Controller (LFLC) tool supplemented with the use of standard XML document when describing the set of fuzzy rules.
Web application for editing linguistic description for LFLC and the following communication with the LFLC fool from the web client (entering input data, starting LFLC with the linguistic description, visualisation of outputs from the LFLC tool).
The methodology of the application of F-transformation for the analysis of experimental data from the system, verified on data from real objects in order to be used in signal processors.
Implementation of high-interaction honeypot including assessment of the results.
The methodology of attractivity improvement of low-interaction honeypots for automatised attackers including continuous assessment of attack samples.
Adaptive e-learning system controlled by the hierarchical state and several fuzzy expert systems.
Methodology of the scientific and research performance will be based on the architecture of a general tool for fuzzy modeling, starting from a general model for decision making under uncertainty, and from the previous research.
Adaptive differential evolution algorithms and their applications
The research aims at studying adaptations in differential evolution algorithms, and solving the issues of stagnation, parallel algorithms and implementation of methods. Our objective is to publish a book of new findings acquired at conferences and in scientific journals, and if the case may be to extend publicly available libraries of global optimisation programmes.
Modeling of operating processes
The research focuses on the Design Engineering Methodology for Organisation (DEMO), which in based in the “Enterprise Ontology“ theory and which represents a generic approach to operating processes modeling. It further focuses on value modeling of operating processes that monitors the changes in values of economic resources, and that is domain specific. The aim of the research is both practical application of DEMO methodology and shared use of both methodologies, which would enable better and more efficient processing of operating processes value modeling.
Updated: 08. 11. 2017