Relazione su invito
Astronomical Big Data: The template case of photometric redshifts.
Modern and future digital sky surveys will produce data in the multiTera and Petabyte regime. These data are both heterogeneous and complex with many hundreds of parameters measured for each of the billion objects which are detected. Their reduction, analysis and interpretation pose brand new problems which require the adoption of automatic methods largely derived from the Artificial Intelligence and Machine Learning domains. Among the many problems which are currently under investigation, feature selection $(i.e.$ the identification of the most significant parameters to be used for a specific task), the merging of different methods in a coherent way and the proper evaluation of errors are the most crucial. Photometric redshifts which are crucial to provide an estimate of the distance of the large samples of galaxies required by modern precision cosmology are a great showcase of the above. We shall discuss in detail the application of machine learning methods to the derivation of redshifts in the context of the KiDS (Kilo Degree Survey), VST-VOICE and other surveys.