SYMBOLIC REGRESSION OF DETERMINISTIC CHAOS

Presented paper deals with problem of chaotic system identification and description using Genetic Programming Algorithm. To prevent random perturbations caused complicated identification of regression model parameters, hybrid GPA-ES model is applied. As a test case, Lorenz attractor data are used and this model is restored from data by above mentioned GPA-ES model. During the work, interesting influence of fitness function computation strategy was identified. The paper concludes by comparison of two possible strategies computation of data in each step of time series when original data are used on the place of initial state and continuous computation of whole time series from single starting point.