SB/BNL Joint Cosmo seminar: Boris Leistedt, NYU: "Forward, causal modeling of galaxy photometry (or Hierarchical causal models with embedded machine learning)

Thursday 15 Nov 2018, 15:00 → 16:00 US/Eastern

I will describe a new self-calibration approach based on hierarchical modeling that delivers precise redshifts as well as recalibrated photometry and SED models for galaxy surveys. I will show a demonstration of this approach on early DES data, and how it opens new perspectives for the exploitation of ongoing and upcoming surveys such as LSST. I will also discuss the use of machine learning in physics, and two exciting new avenues of research: encoding physics in machine learning algorithms, and embedding machine learning in hierarchical causal models.