Stochastic search-based neural networks learning algorithms

We report on the application of artificial neural networks (ANN) learning algorithms, based on stochastic search, in learning and optimization processes. In our earlier studies we have shown a possibility for the application of the stochastic search algorithm (SSA) in relation to system optimization and identification, as well as learning processes of certain types of ANN. Here we modify stochastic search method by using certain algorithms as an alternative in ANN learning process. Furthermore, we compare the results from SSA and back propagation error (BPE). In certain cases, SSA are more favourable vs. BPE, particularly during complex learning processes on static ANN. Thus, SSA is effective engineering tool in ANN optimization, as well as in learning processes. We have tested different SSA examples with our specially developed software application.