Multi-Valued Neuron with a periodic activation function - as part of a multi-layered Neural Network

The increased functionality of Multi-Value Neuron with a periodic activation function (MVN-P) lead to the idea of integrating it into a multi-layered Neural Network (MLNN) to increase even more its capabilities. For such MLNN up to 1 MVN-P, as output neuron, is sufficient in order to increase the over roll efficiency. Back-propagation correction rule can be applied in order to correct the weights, leading to NN architectures close to that of a classical feed-forward NN. Nonlinear multi-thresholds problems were used through-out the validation process with encouraging results presented in the current paper.