The selection of spread and the application in engineering of GRNN

The algorithm of the generalized regression neutral network (GRNN) is introduced. Compares and analysis with superiorities and characteristics between GRNN and BPNN. The effects and choices of spread are discussed. It is very important to confirm the value of spread. Appropriate value of spread is searched out from all specimens based on the LOO across validate method. A simple example is given to prove the accurate value of spread can improve the generalization ability of the network. The method of GRNN is applied to the examples of slope stability evaluation, it is concluded that GRNN modeling is simpler and more accurate than BPNN in prediction and analysis of slope stability. Artificial neural network algorithm has been widely used in engineering, and GRNN will be accepted for engineers more easily as a means of more advanced algorithms. It will play an increasingly important role in the non-linear calculations.