Extracting categories with prototypes in artificial cognitive agents

In this paper a general strategy of determining categories with prototypes is presented and its variations are discussed. It is assumed that the task is carried out by an artificial agent which autonomously develops and maintains its private ontological knowledge base. The computation of a category and its prototype is based on a learning set consisting of messages obtained by the agent from other participants of external communication processes who are considered teachers and treated as sources of new meanings. The teachers communicate their beliefs related to an inclusion of particular objects to a category which the listening agent is trying to learn. Potential categories are defined over a related cognitive space defined with respect to a particular distance or similarity measure, both available to the artificial agent along with computational mechanisms for determining central objects in learning sets. Simplified computational examples of calculations performed within the proposed strategy are presented.