Detail publikace

Learning methods in expert systems

VALENTA, J.

Originální název

Learning methods in expert systems

Typ

přednáška

Jazyk

angličtina

Originální abstrakt

This presentation describes learning methods used in expert systems. The motivation for learning knowledge-bases in expert systems is there presented, as well as the basic of an expert system. Presentation introduced a method to eliminate knowledge engineer from the process of expert system's knowledge base creation. The methods developed for this task are mainly based on fusion of neural networks and rules. Two basic algorithms are shortly presented here and their advantages and shortcomings are noticed. These methods are compared with two developed methods for the direct learning of rule-based systems. One of the methods consists on a modification of back-propagation algorithm to learn the rule base directly and the other is based on evolutionary computation with special uncertainty processing. The structure, properties and method of learning are presented, as well as the process of using the algorithm and its main advantages. Finally is introduced an implementation of the learning algorithms in a real expert system.

Autoři

VALENTA, J.

Vydáno

5. 12. 2007

Místo

Coimbra, Portugalsko

BibTex

@misc{BUT64768,
  author="Jan {Valenta}",
  title="Learning methods in expert systems",
  year="2007",
  address="Coimbra, Portugalsko",
  note="lecture"
}