Simulation results of the Wang-Mendel algorithm combined with the technique of cross-validation suggest an extension of this algorithm that-within a very wide class of possible Fuzzy Associative Memory (FAM) banks-automatically leads to the simplest FAM-based fuzzy system from a given set of (possibly noisy) numerical data. The new algorithm is tested on a standard problem from the area of control. We also discuss some applications concerning fuzzy data mining and propose conclusive refinements of the new algorithm.
展开▼
