Memory-oriented neural network with indication of units assigned to the inputs

摘要

The invention relates to a memory-oriented neural network with indication of units assigned to the inputs, a method of adaptively dimensioning this network and a method of classifying patterns fed to this neural network on the input side. The neural network is formed by a multiplicity of units representing neurons. The units are provided by memory locations of a memory device. The connections in the neural network are realised by indicators providing addresses. The indicators, to which if need be positive weightings are assigned, are directed from the units assigned to the inputs of the neural network in the direction of the outputs of the neural network. <IMAGE>