A self-organizing neural network for cluster detection and labeling

We present an artificial neural network, which based on a given generic interpoint similarity measure is capable of clustering a set of data, and then assigning to each new input its appropriate cluster label. The network has been called a cluster detection and labeling (CDL) network. It consists of two layers. The first layer is a `similarity-measuring´ layer, which calculates the similarity of a new input pattern with representatives (prototypes) of clusters stored in the network. The second layer of the network assigns a cluster label to each new input pattern. We give a brief description of the network structure and algorithm, and show the performance on clustering some artificially created data sets