Theoretical results on applying EM-SFM algorithm and modular network to fuzzy Sugeno model (Q2745586)

The study re-examines the well-known Takagi-Sugeno fuzzy model in the setting of modular networks (where it is noted that each expert network of a modular network represents a conclusion part of the rule of the fuzzy model). Then a complete EM algorithm (iterative determination of maximum likelihood estimates) is discussed along with a study of convergence properties of the approach. It is shown that, under some assumptions, the EM algorithm applied to this estimation problem exhibits linear convergence. A simple numerical example is included.