Direction-basis-function neural networks
Wang Shoujue, Shi Jingpu, Chuan Chen, Yujian Li · 2003
A model called direction-basis-function (DBF) neural networks and its relevant algorithm for pattern classification are proposed. Also, one implementation of the model based on the architecture of priority ordered neural networks (PONN), that is PODBFN, is discussed. When adapted as pattern classifier, the neurons with linear output are used in the course of learning, while hard limited step activation function is for pattern classification. The computer simulations show that not only the convergence speed is much faster than the improved BP algorithm, but also the performance of the network is better. The DBF pattern classifier was implemented on the micro-neural computer CASSANDRA-I, and the experimental results obtained are presented.