Metallicity is a fundamental parameter in the study of galactic evolution, providing us with information on the object’s age and on its star formation history; together with Initial Mass Function (IMF), age and Star Formation Rate (SFR), metallicity allows us to fully constrain a galaxy Spectral Energy Distribution (SED), which is the theoretical basis of modern evolutionary population synthesis models.
However, because of the difficulties in observing reliable spectral features outside nearby galaxies and of the advantages presented by photometric observations over spectroscopic ones (in terms of telescope time efficiency and of the possibility to observe fainter objects), it appears suitable to develop ways to determine galactic metallicities through the observation of photometric properties alone.
In this sense, the aim of this work is to present the application of Machine Learning techniques to the problem of photometric metallicity determination; using datasets of galaxies of known metal content, we trained a Multi- Layer Perceptron (MLP) neural network with a Quasi-Newton Algorithm (QNA) as learning rule, in order to make the algorithm able to derive metallicity from observable photometric quantities.