Research of Learning Performance Based on the Area Division Forward Competition Neural Network
XU Chen-xi ,LI Yan-ren (Department of Radar System Engineering, AFRA,Wuhan 430019, China)
In this paper we analyze the shortcomings of the competitive neural network and area division com-petition neural network (ADCNN), present an improved learning law, Kohonen rule based on the soft-boundary pro-cessing in ADCNN. Combined with ADCNN and soft-boundary processing, it can deduce training specimen and im-prove learning speed. It can be used for the waveform identification of analog signals through features of waves.
【CateGory Index】： TP181