The use of the RCE Neural Network in a Pattern Recognition

A Restricted Coulomb Energy (RCE) neural network is a specific design of the hypespherical classifier proposed by Reilly in 1982 and it is suitable for classification of nonlinearly separable clusters of patterns in multidimensional pattern spaces. The paper deals with the use of the RCE neural network for the recognition of persons. Two sets of input patterns were used in described experiments. The first set contained pictures of person faces and the second set contained person fingerprints. The short principle and topology of the RCE neural network, learning and retrieving algorithms and mentioned experiments are described in the paper.