Day 2

Nice, France.  Day 2 of vacation and astroph blog.


1105.3975
Photometric redshifts and quasar probabilities from a single, data-driven generative model
Bovy, Meyers, Hennawi, Hogg, etal.


* Are they talking about spectra or multi-band photometry classification?  I guess normally it would be from spectra, but I'm preoccupied with multi-band photometry and so easily confused.


A new technique: simultaneously classifies and estimate z of quasars; incorporates flux uncertainties, missing data, multi-wavelength photometry (is spectra given assumed??), from models of quasars ("extreme deconvolution technique") in flux-redshift space, to estimate the underlying density.  This allows for quasar flux densities to be estimated over a chosen redshift range.  Speed obtained by combining using analytical forms for the basis, and using many simple components.  This method is competitive with the best photometric techniques and photo-z techniques.  Inclusion of UV and NIR data significantly improves quasar-star separation (they were talking about ugriz system!)  by a factor of 3.  Code publicly available.


* This abstract was hard to decipher to a non-expert like myself, mostly because the basic assumptions are not described.