The paper deals with global property analysis of the Simultaneous Perturbation Stochastic Approximation algorithm. It spite of the fact that the algorithm uses an estimation of a gradient, a global optimisation property can be achieved by using properly shaped sequence {ck} used for generating trial points. The experiments are performed on a number of test functions showing efficiency of the method. The examined algorithm is also used to train a dynamic neural network. A network under consideration is designed using neuron models with internal dynamice. The identification of an industrial process based on real data shows usefullness of the algorithm.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.