Skip to main content

A Self-Adaptive Fuzzy System that Improves its Performance over Time

Partridge T, Tjahjadi T

Proc. 2nd IEEE Int. Conf. on Fuzzy Systems, 1993, 506-511

Abstract

This paper presents a machine learning paradigm which fine-tunes and improves a system of fuzzy implications. The method uses connection matrices and experts' credibility weigths to combine several rule-sets into one knowledge base. The results of each expert system decision are fed back and used to adapt the weights and, at a finer grain, the rules of the expert system. The reasoning process are evaluated and adapted to deal with new facts. Hence, the control structures are fine-tuned to deal with the real world. The resulting expert system is an open system that uses frames, rules, fuzzy implication and connection matrices to produce a form of machine learning. This machine learning is defined as a self-adaptive fuzzy system that improves its performance over time. This method is implemented in an expert system for X-ray rocking curve analysis.