Statistical Challenges in photometric redshift inference

Observations from large area photometric surveys like LSST or DES will constrain cosmology to unprecedented precision. Deep wide-area imaging will provide observations for faint galaxy samples, for which traditional redshift calibration using spectroscopic data is very difficult. It is therefore important to quantify and incorporate the modelling uncertainty in the parametrization of `systematics' like photometric redshifts or galaxy-dark matter bias in Large-Scale Structure and Weak Lensing probes.

I will present our recent work on a Bayesian model combination approach that allows both a joint inference of these systematics, as well as a consistent treatment of their respective modelling biases. I will particularly discuss how the statistical inference can be scaled to the large galaxy samples in ongoing and future photometric surveys and incorporated into the current cosmological inference methodology.

Speaker: Markus Michael Rau, Carnegie Mellon University