Multiple-Valued and Branching Neural Networks

The generalizations of common single-valued artificial neural networks are proposed. The main proposition is to consider multiple-valued networks. Such neural networks can have multiple values of elements and multivalued connections between elements at given time moment. Modified Hopfield neural networks with strong anticipation property are considered as the examples. New aspects of learning processes considered for the case of multiple-valuedness. Branching networks are described. Presumable applications of multivalued neural networks are proposed. Also some new research problems are described.