Abstract:

Based on spline approximation theory and neural networks, a spline neural network is constructed, where the hiddenlayer neurons are activated by a group of spline functions. Based on the standard error backpropagation (BP) method, the neuralweights updating formula is derived. More importantly, a pseudoinverse based method is then proposed, which could directly determine the network weights without iterative training. Computer simulation results show that the onestep weightsdetermination method could be more effective than the standard BP iterativetraining method.