Immunological selection mechanism in agent-based evolutionary computation

  

Aleksander Byrski
AGH University of Science and Technology in Cracow, Poland

For some problems applying of classical optimization techniques seems to be ineffective, and heuristic approaches must be applied (e.g. evolutionary algorithms which are popular and universal). Evolutionary multi agent-systems (EMAS) which consist in combining the biological inspirations (evolutionary algorithms) and social ones (agent-based systems) are interesting optimization technique researched at the AGH University of Science and Technology in Cracow, Poland.

In this talk the introduction of immunologically inspired selection mechanism into evolutionary multi-agent systems is presented. This mechanism should help in lowering computational complexity when the problem to be solved by evolutionary approach requires the computation of a very complex fitness function (e.g. evolution of neural-network architecture). After presentation of the concept of the EMAS and immunological EMAS and basics of the formal model of the system, experimental results concerning the optimization of the multi-modal multi-dimensional benchmark functions will be shown. Then the concept of evolutionary-neural agent-based evolutionary predicting system will be shown along with some experimental results.


Aleksander Byrski is assistant professor at the Department of Computer Science, AGH University of Science and Technology in Cracow, Poland. He works at the Intelligent Information Systems Group led by Prof. Edward Nawarecki. He holds M.Sc. and Ph.D. in Computer Science. His main research interests are in the area of biologically-inspired computer systems (e.g. agent-based computation, immunological and evolutionary algorithms). He teaches several courses in computer science, i.a. evolutionary algorithms, object and component technology, neural networks.