Machine learning methods can reveal the relations between observational data and the properties of the observed objects. They simultaneously have the potential to complement and inform the build up of physical models, provide data-driven predictions for practical applications, and automatize the analysis and comparisons of large data sets (real or simulated).

A number of machine learning tools have already been proven successful in a variety of astrophysical contexts, e.g. from the measurement of galaxy morphologies to automatic object detection and identification (e.g. quasars in the early Universe or RR Lyrae stars around the Milky Way), streamlined estimation of galaxies’ redshifts, and image analysis techniques for de-noising and deconvolution. In the upcoming years efficient machine learning and image analysis algorithms will become of the essences to fully exploit the available data and make substantial progress in our understanding of astrophysical phenomena.

We are therefore planning a hands-on and interactive workshop to foster big-data and machine-learning proficiency in the astronomy community, with particular focus on applications, among others, pertaining extra-galactic and galactic astronomy, galaxy formation, large scale structure, cosmology and cosmological surveys. Tools will range from classical regression and classification methods to state-of-the-art deep learning algorithms as generative adversarial networks and variational autoencoders. We plan to invite to participate also a few representatives of big and small private companies, in order to favor the percolation of their know-how expertise to the astronomy community.