Agglomerative learning for general fuzzy min-max neural network

Abstract

Proposes an agglomerative learning algorithm based on similarity measures defined for hyperbox fuzzy sets. It is presented in a context of clustering and classification problems that are tackled using a general fuzzy min-max (GFMM) neural network. The agglomerative scheme's robust behaviour in the presence of noise and outliers and its insensitivity to the order of the training pattern presentation are used as a complementary features to an incremental learning scheme, making it more suitable for online adaptation and dealing with large training data sets.