Experimental studies with a hybrid model of unsupervised neural networks

This paper presents an unsupervised clustering method to classify the optimal number of clusters from a given dataset based solely on the image characteristics. The proposed method contains a feature based on the hybridization of two unsupervised neural networks, Self-Organizing Maps (SOMs) and Fuzzy Adaptive Resonance Theory (ART), which has a seamless mapping procedure comprising the following two steps. First, based on the similarity of the spatial topological structure of images, we will form a local neighborhood region holding the order of topological changes. Then the region is mapped to one-dimensional space equivalent to more than the optimal number of clusters. Furthermore, by additional learning in accordance with the order of the one-dimensional maps formed in the neighborhood region, we must generate suitable labels that match the optimal number of clusters. We use it as a target problem for which the number of categories or clusters is unknown. We emphasize the effectiveness of the proposed method for resolving the target problem for which the number of categories and clusters is unknown, and we anticipate its use for the categorization of facial expression patterns for time-series datasets and for the segmentation of brain tissues shown in Magnetic Resonance (MR) images.