FURL - A theory revision approach to learning fuzzy rules

Fuzzy Rules have been shown to be very useful in modeling relationships between variables that have a high degree of uncertainty or ambiguity. A major question in regards to learning fuzzy rule bases is how to handle interactions between rules of overlapping coverage. Structures, such as Yager's HPS (Hierarchical Prioritized Structure), have been proposed to answer this question. In this paper, we present our system of learning a hierarchical fuzzy rule base named FURL (for FUzzy Rule Learner). FURL takes advantage of the properties of HPS to learn hierarchical levels of fuzzy rules. FURL applies machine-learning techniques from theory revision such as credit assignment and repair to fuzzy rules. We also discuss the results of testing FURL on multiple benchmark datasets and finally discuss our results.

FURL - A theory revision approach to learning fuzzy rules Conference Paper