A novel neuron model and its application to classification

This paper presents two new neuron models with the application to classification. The models, namely RRN (neuron based on residue reduction), and RNSN (neuron based on residue number system), are similar in that all the arithmetic operations are confined in the ring of integers module M(Z_M). Their processing units are identical, that is computing the remainder of its total input. Their inputs are different: RRN uses the traditional binary or decimal representation, while RNSN uses the residue number system, and hence makes it more flexible. Both RRN and RNSN are more capable in classification than perceptron, they can realize many linearly inseparable functions, such as the XOR problem. The difference between perceptron and the neuron models is discussed