Astronomy in the 21st century: Drowning in data, Starving for knowledge
par
Salle 9110
For a long time, Astronomy was considered a data starving science. The technological difficulties associated with the observational process lead the field to evolve around the study of a few, well observed, objects. This paradigm has shifted completely in the last two or three decades with the advent of large scale surveys. In this new scenario, traditional approaches to data analysis in a case by case basis are no longer feasible. Astronomy is reinventing itself through interdisciplinary collaborations with computer science and statistics. In this talk, I will briefly describe how this partnership developed - from the earlier applications of Artificial Neural Networks in the early 1990's to the more contemporary discussions around Domain Adaptation and Active Learning. Using the determination of photometric redshifts as a case study, I will discuss a few algorithms, the scientific outcomes they enabled and the challenges still lying ahead.
