DARWIN: An evolutionary program for nonlinear modeling of chaotic time series

DARWIN is an efficient evolutionary algorithm programmed to approximate the functional relation, in symbolic form, that describes the behaviour of a t ime series. The search procedure is based on Darwinian theories of natural selection and survival. An initial population of potential solutions is sub jected to an evolutionary process described by selection, reproduction and mutation processes which are repeated over generations until an optimum ind ividual is finally found. DARWIN is particularly useful when the dynamical model that creates the time series is nonlinear. The code, based on a previ ously proposed evolutionary algorithm [Phys. Rev. E 55 (1997) 2557-2568], i s programmed in Fortran 77. Darwin is actually employed as a predictor in a satellite based ocean forecasting system. (C) 2001 Elsevier Science B.V. A ll rights reserved.