Abstract
Fuzzy models can be conveniently regarded as linguistic modelling structures with well-defined functional blocks of input and output interfaces along with a processing module. This paper examines the functions of these modules and specifies the relevant optimization tasks which accrue to them. Considering several levels of memorization completed within fuzzy models (resulting in establishing short-, medium- and long-term memories), the corresponding learning policies are developed. Some classes of optimization mechanisms are also discussed, with particular emphasis focused on their ability to handle multimodal problems. These tools comprise both gradient-based techniques and selected methods of evolutionary optimization, in particular genetic algorithms. Illustrative numerical studies are also included.