Application of likelihood search method to neural networks learning

This article mainly compares learning performances between the Likelihood Search Method (L.S.M.) and the Back Propagation Method (B.P.). The performances are evaluated by the simulations which include both static and dynamic Neural Networks (N.N.) learning problems. In the simulations, N.N. is trained to realize nonlinear functions or control a nonlinear crane system by using the L.S.M. and B.P.. Simulation results show that the L.S.M. is superior to the B.P. because of the ability of intensification and diversification of the search.