Coffee talks
Friday 19/07/2019 @ 11:30, Sala riunioni quarto piano
Claudio Gheller (Swiss Plasma Center, EPFL Switzerland), "Exploring High Performance Machine Learning Solutions for Radioastronomy"
Machine Learning (ML) has gained increasing popularity in the last decade thanks to the concurrent availability of enough data and computing power necessary for its effective usage. ML based solutions promise high accuracy, great speed and full automation (no need of human supervision) in a broad spectrum of applications (some of them already successfully adopted in common devices like smartphones or in cars). Accuracy, speed and automation are key features for (radio)astronomical data, especially having to deal with the large volume of rich and complex data that is being delivered by instruments like LOFAR, MWA, ASKAP, MeerKAT and that will dramatically grow in the perspective of the SKA. We are exploring the possible application of ML solutions to radio observations, focusing, in particular, on source detection and classification and image denoising. The essential architectural and functional characteristics of our ML solutions will be introduced. The current achievements and results will be discussed, highlighting advantages strengths but also drawbacks and limits of ML solutions compared to standard approaches (for a first application, see Gheller, C.; Vazza, F.; Bonafede, A., 2018, MNRAS, 480, Issue 3, p.3749-3761, arXiv:1809.03315).