Evolution strategies applied to the problem of line profile decomposition in QSO spectra.pdfVIP

To appear in Astron. Astrophys. Evolution strategies applied to the problem of line profile decomposition in QSO spectra R. Quast, R. Baade, and D. Reimers Hamburger Sternwarte, Universit¨at Hamburg, Gojenbergsweg 112, D-21029 Hamburg, Germany 4 e-mail: {rbaade,dreimers}@hs.uni-hamburg.de 0 0 Received July 6, 2004 / Accepted September 16, 2004 2 t Abstract. We describe the decomposition of QSO absorption line ensembles applying an evolutionary forward modelling c O technique. The modelling is optimized using an evolution strategy (ES) based on a novel concept of completely derandomized self-adaption. The algorithm is described in detail. Its global optimization performance in decomposing a series of simulated 8 test spectra is compared to that of classical deterministic algorithms. Our comparison demonstrates that the ES is a highly 1 competitive algorithm capable to calculate the optimal decomposition without requiring any particular initialization. 2 Key words. methods: data analysis – methods: numerical – quasars: absorption lines v 1 9 3 1. Introduction mum of f crucially depends on the adequacy of the initial guess 0 x(0) . Hence, the results are biased and several optimization runs The standard astronomical data analysis packages such as the 1 are normally necessary. In particular in case of noisy data or 4 Image Reduction and Analysis Facility (IRAF) and many pop- strong inter-parameter correlations the problem of finding a