Orthodoxy basis based on procedure neural networks

This paper proposes several approximations and algorithm issues in procedure neural networks (PNNs). In the PNNs the weights are associated with time and can be represented by certain basis functions. The choice of weight functions affects the property of PNNs, especially in training of the PNNs. Orthodoxy basis functions have many advances in representing the weights and saving time in PNNs learning. In this paper several kinds of orthodoxy functions are proposed and the corresponding experiments support these works. However convergence in training the PNNs is another important issue in analyzing the property of PNNs, and this paper discusses some related problems in a learning algorithm.