On genetic representation of high dimensional fuzzy systems

Abstract

We explore evolutionary design of fuzzy systems for high dimensional input systems. We evaluate three fuzzy system representations that avoid the exponential increase in the number of rules as the input dimension increases. We report on system design aspects such as the performance of the search algorithm, and the quality and complexity of the final solution. Based on our results, we discuss the merits and drawbacks of each representation and propose a technique to mix representations.