Galaxy Evolution with Machine Learning

The study of the evolution of galaxy properties with redshift is essential for constraining how galaxies form and evolve, and the development of efficient tools for this task is necessary to deal with the large volumes of high quality data provided by current and future surveys. In this talk I will summarize two projects which we are developing at São Paulo. In the first case I will show how to introduce emission lines in simulated galaxy spectra (which in general present only the continuum and absorption lines) in order to produce realistic spectra for mock catalogues and tests of the data reduction pipelines of coming surveys, like PFS and J-PAS. I will also show how to obtain stellar population parameters (e.g. mean ages, metallicities, reddening) from optical spectra through a Bayesian analysis of the observations with a training set of simulated theoretical spectra.