Data: 05.05.2016

Palestrante: Dr. William Hartley (University College London)

Link: https://youtu.be/exmHzIKhZhU

Resumo: There are a number of ambitious ongoing and forthcoming cosmological experiments utilizing information contained within multiple broadband images of galaxies. Among the most important measurable quantities in these surveys are the distances to the galaxies used for a weak lensing or BAO analysis via their photometric redshift. Two approaches are typically taken to the problem of deriving photo-z: modelling the galaxy population and using machine learning techniques to directly map from the photometry to the likely redshift. Machine learning methods are proving to be the more powerful presently, but face fundamental difficulties in the future due to the lack of spectroscopic information. It is vital, then, that modelling methods become competitive. In this webinar I will outline some of the major challenges and necessary developments in pursuit of this goal.