Layered MVL neural networks capable of recognizing translated characters

摘要

The multivalued logic (MVL) neurons constituting the layered MVL neural network use MVL operations to produce analog responses to be fed to the respective quantizers. A four-layered MVL neural network model capable of recognizing translated characters is presented. Translation of input characters is easily carried out because MVL neural networks have the unique ability that the input patterns for which such a network has been trained can be reproduced directly from the states of synapse weights. Simulation results showing successful recognition of translated characters are presented.