A novel algorithm of incremental construction of nonlinear parametric approximators is introduced that needs a relatively small number of parameters to achieve good accuracy. The approximator is defined as a linear combination of nonlinear base functions, and the method adds the base functions one by one, observing the correlation between the residue function and the candidate for the base function. The base functions do not need to be homogenous, i.e. to share the same formula. The method was experimentally verified and compared to neural networks, and the preliminary results were encouraging enough to develop the presented approach further.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.