A Modified FMM Neural Network for Pattern Classification and Feature Analysis

H.J. Kim and H.S. Yang (Korea)


Pattern classification, FMM neural networks, featureanalysis


In this paper we introduce a modified Fuzzy Min-Max (FMM) neural network model for pattern classification and feature analysis. The proposed model employs a new activation function which has the factors of feature value distribution and the weight value for each feature in a hyperbox. During the learning process, the feature distribution information is utilized to compensate the hyperbox distortion which may be caused by eliminating the overlapping area of hyperboxes in the contraction process. The weight factor makes it possible to consider the degree of relevance of each feature to a class during the classification process as well as the learning process. We define a new hypercube membership function and a learning method for the proposed model such as hyperbox creation, expansion, contraction and weight updating scheme. The proposed model also can be applied to feature analysis process in a pattern classification problem. For the purpose, we define a Relevance Factor that represents a degree of relevance of a feature to the given class and a similarity measure between fuzzy membership functions of the hyperboxes.

Important Links:

Go Back