Abstract: the principle and algorithm of neural networks of Radial Basis Function (R BF) and its application to the expert system for earthquake prediction (ESEP 3.0) are introduced in this paper. The actual application in earthquake forecast shows that the neutral networks can overcome some demerit of BP neural networks in leaning process, the constringency excessively depend on initial value and optimization constringency and often can′t appear. The RBF neural networks possess the rapid operation speed in learning and strong nonlinear mapping ability and very good efficiency.

Key words: Neural networks of radial basis function, BP neural networks, Learning method, Expert system