Hits:

Indexed by:会议论文

Date of Publication:2008-01-01

Included Journals:CPCI-S

Page Number:590-595

Key Words:Fuzzy neural network; Max-min fuzzy neural network; Error function; Lattice degree of nearness

Abstract:The error function of a fuzzy neural network (FNN) is usually defined as the sum of the squares of the differences between the desired outputs and the actual outputs for all the patterns to be learnt. This error function can be viewed as "distance" nearness between two fuzzy sets, and is a type of similarity measure. In this paper, a new definition of the error function and a corresponding new training algorithm are proposed for max-min FNNs in terms of the lattice degree of nearness, which is a type of "shape" similarity measure between two fuzzy sets. As demonstrated by two numerical experiments, the new algorithm shows better convergence than the conventional one.