To view the structure and parameters of a BP neural network from the knowledge

摘要

The neural network has excellent learning ability, but the weights and the thresholds of it can't be explained and understood, which brings difficulty to the further applications. In this paper, from a different angle, an idea to view the structure and the parameters of a general BP neural network from the knowledge is proposed. Using the knowledge obtained form the samples, the size and the parameters of the network can be determined. The feature of this method is that the structure and the traditional BP learning algorithm of the network is not to be changed any more. The network constructed in this way can approximately track the output without learning, so the learning time of the network can be reduced. Meanwhile, the significance of the weights and the thresholds can be explained. The simulation result proves the validity of this method.