NetFAN - A Structured Adaptive Fuzzy Approach

Adaptive fuzzy systems are usefull universal function approximators. But they suffer from the curse of dimensionality. I.e. the number of parameters, which have to be tuned, increases drastically if the number of input variables increases. This has the effect that memory and computational demands increase also drastically, and more stringently fitting problems may occur if the number of training data is limited. The approach presented in this paper addresses both problems by decomposing the functional mapping into a Network of Fuzzy Adaptive Nodes (NetFAN). This decomposition reduces the number of parameters as well as memory and computational demands. Some first investigations outline the basic characteristics of the NetFAN-approach.