Coffee talks
Friday 10/05/2024 @ 11:30, Sala riunioni quarto piano e on-line (meet.google.com/sue-bwvk-axf)
Ivano Baronchelli (INAF-IRA Bologna), "NOISEMPIRE, an empirical approach to ALMA noise simulation"
In the near future, interferometric imaging is going to become slow and difficult due to the huge amount of data that next-generation telescopes like ALMA (after the upcoming wideband sensitivity upgrade, or WSU) and SKA will produce. Machine learning algorithms could be the answer to this problem, but training them needs robust, denoised ground truth data for comparison. Existing simulations, based on theoretical models of noise, don't capture the full picture, especially some of the unpredictable noise patterns unique to interferometry. We propose a new empirical approach to capture and simulate these overlooked noise patterns. This will create more realistic training data for ML algorithms in interferometry, speeding up and improving image processing.