Paper 13101-175
A modular machine learning data processing pipeline for astronomical observatories
Abstract
This project introduces a new Machine Learning (ML) pipeline aimed at enhancing data processing in astronomical observatories. This automated pipeline aims to significantly reduce nightly data processing time and false positives rate. By integrating various deep learning algorithms and adhering to ML Operations (MLOps) principles, the pipeline promises enhanced performance and adaptability for present and future observatories. It will streamline the analysis of astronomical data, improve transient detection, and allow for more efficient sky surveys, enabling the study of a wider range of celestial phenomena. The modular design of the pipeline not only boosts efficiency but will also foster collaboration and innovation within the astronomical community.
Presenter
Fiorenzo Stoppa
Radboud Univ. Nijmegen (Netherlands)
Dr. Fiorenzo (Fiore) Stoppa is a Postdoctoral Researcher at Radboud University, specializing in astrostatistics and astroinformatics. He holds a Master's degree in Statistics from the University of Bologna, where he also completed his Bachelor's studies. His research integrates statistical methods and machine learning to enhance data analysis in astronomy.
Currently involved in the BlackGEM telescope array located at ESO's La Silla Observatory in Chile, Dr Stoppa focuses on developing and applying novel machine learning methods for source localization, feature extraction, classification and transient detection.
His current interest is the development of a unified, end-to-end astronomical data processing pipeline that effectively integrates multiple deep learning algorithms, adheres to solid ML Operations (MLOps) principles and aims at continuous performance enhancement over time.
He is more than happy to discuss about astroinformatics in English, Italian and basic Japanese!