Evolution strategies applied to least-squares curve fitting of spectroscopic data

A robust method for nonlinear least-squares curve fitting without using function derivatives is presented. An advanced evolution strategy is applied to the fitting of high-resolution absorption spectra of polyatomic molecules to estimate half-widths and broadening coefficients. A comparison with the Levenberg-Marquardt algorithm shows that the evolution strategy yields the same results while being less sensitive to the choice of the initial parameters, particularly in the case of crowded spectra. Copyright (C) 1996 Elsevier Science Ltd