IDENTIFICATION OF SYSTEMS DESCRIBED BY RELATIONAL KNOWLEDGE REPRESENTATION

摘要

The paper is concerned with the identification of a static system described by a relational knowledge representation with unknown parameters. The identification consists in step by step knowledge validation and updating using the results of observations, and may be considered as an extension of the parameter estimation for a system described by a traditional mathematical model. Two versions of the identification algorithm have been presented for two different possibilities of the observations. The applications of the identification algorithms in the learning decision making (control) system and in a problem of knowledge-based recognition have been described. In the second case the knowledge representation consists of relations concerning the features describing an object to be recognized and the index of a class to which the object belongs. Related problems have been discussed: identification with a logical knowledge representation and with uncertain observations, application of uncertain variables, identification of a system with distributed knowledge.