Identifying fuzzy rules

Abstract

In Chapters 6 and 8 we discussed separately symbolic rules for identifying membership functions to cluster and subsymbolic rules to learn functions that suitably translate a set of inputs into an output variable. Here we look for complete procedures performing both tasks having the final aim of computing suitable functions from input to output. The general direction is forming these functions as a composition of information granules with mixing functions. Hence we look for a corresponding composition of cluster identification with function learning methods, possibly producing additional benefits in respect to the union of the advantages of the separate methods. Within a wealth shell where any method is admissible provided it is based on meaningful properties of granules and feasible computations, here we focus on three families of procedures: i) the training of a single granular neuron; ii) the partially supervised identification of fuzzy clusters; and iii) the overall training of a neuro-fuzzy system. © 2008 Springer-Verlag Berlin Heidelberg.