Intelligent System in Some Decision Classes' Environment

Abstract

One of the trends in the information theory of recent years is a dynamic development of decentralized, intelligent information systems. The systems in question are so complex that the classical methods of describing their dynamics, prove to be insufficient. One of the possible ways of describing them is by applying, the model of an agent which constitutes an independent entity existing in its own right. The agent has its own inventory of objectives and the ability to interact with the environment and adapt to the changes which take place in this environment. The paper presents the conception of autonomous agent as a cybcrnetic model of an autonomous system. lt is based on psychological and biological phenomena which may be observed in living organisms. The description of the system distinguishes between three basic functional blocks: correlator, homeostasis and accumulator; it contains a detailed account of the functions and mutual correlations and interactions. Subsequently, the paper presents a classification of interactions between correlator and homeostatis, as well as between homeostasis and accumulator. An analysis of the autonomous system from the point of view of the psychological and biological phenomena which take place in it, has allowed one to borrow certain educational methods from beha vioural psychology (clusical conditioning, instrumental conditioning) and apply them to the teaching ofthese type of systems. The paper presents the implementation of these type of systems in artificial neuron networks. Subsequently, the authors show how in practical terms these type of systems may be taught by means of the above mentioned methods. Lastly, the paper presents the application of the above systems to the leaming process together with the discussion ofthe accompanying phenomena.

Keywords

• Multi-agent systems