Neural Fuzzy Systems
In general, there are two main but apparently separate methodological developments relevant to computational intelligence: fuzzy logic systems and neural networks. Fuzzy logic systems try to emulate human-like reasoning using linguistic expression, whereas neural networks try to emulate the human brain-like learning and storing information on a purely experiential basis. Both the methodologies have been successfully applied in many complex and industrial processes though they experience a deficiency in knowledge acquisition.
The most important considerations in designing fuzzy systems are the construction of the membership functions (MF) and constructing the rule-base and have been a tiring process. The choice of MFs also plays a decisive role in the success of an application. But there is no automated way of constructing the MFs. They are mainly done by trial and error, or by human experts. As is well recognized, rule acquisition has been and continues to be regarded as a bottleneck for implementation of any kind of rule-based system. In most existing applications, fuzzy rules are generated by an expert in the area, especially for systems with only few inputs. With an increasing number of inputs, outputs and linguistic variables, the possible number of rules for the system increases exponentially, which makes it difficult for experts to define a complete set of rules and associated MFs for reasonable system performance. In Chapter 8, ...