Learning agents for cooperative hyperinformation systems

Abstract

A model of learning agents is presented. It is shown how this model will support the incorporation of agents into distributed hypermedia systems. The question of self-organization in multi-agent systems in which agents are actively learning from their experience is addressed. Each agent, endowed with a different connectionist learning capability, behaves in a specific problem domain and is specialized to interact with environments. Each agent encapsulates a specific set of knowledge obtained from a different set of training examples. The issues relating to the composite and heterogeneous agent models are discussed. The hypermedia information system incorporating heterogeneous and composite learning agents is developed. The aim of this application is to develop the distributed intelligent hypermedia information system in an open environment of distributed workstations.<>