Abstract :
The vast majority of the observable space in astrophysics comes as photons. Spectra therefore represent an optimal source of information. In extragalactic astrophysics, the stellar component produces most of the radiant energy in the rest-frame NUV-optical-NIR window, and mainly consists of a superposition of absorption lines overlaid on a continuum. These photons encode the properties of the photospheres of stars, covering a wide and complex range of mass, age and chemical composition. Population synthesis models have allowed us to interpret the observations, mapping stellar and chemical enrichment histories into photo-spectroscopic observables that can be "inverted" to derive constraints on the underlying stellar content of galaxies. In this seminar, I will give an overview of the standard methodology, and discuss the difficulties underlying the analysis of stellar populations from observable data, and explain, from an information theory approach, the caveats of traditional methods aimed at extracting formation histories from spectra. Recent results following a data-driven approach will be presented that allows us to assess the validity of cosmological simulations of galaxy formation, and to interpret the results regarding the physical processes that control galaxy formation.
