Skip to Main Content
In this paper, we exploit the capability of evolutionary programming for construction and training neural networks, independent of the applied models of the neurons. The main application of this algorithm is training neural networks with elaborated models for neurons. For instance when because of implementation limitations a deviation from ideal models is mandatory, this algorithm can be used to take these deviations into account during the training process. The functionality of the proposed algorithm is demonstrated by training a neural controller with nonconventional neurons.