Papers of interest on the arXiv today – Sept 26th 2016
Two interesting papers on the methodology side today: A ML source detection method that detects ultra-faint streaks below the pixel level noise (arXiv:1609.07158). They call it “ML” but in reality it uses MCMC to look at the Bayesian posterior (which is arguably a good thing!) and even Bayesian model comparison (in the BIC variety) to determine how many parameters the streak model should have.
This paper applies a Bayesian method to extract 2M distances to stars in the Gaia DR1 catalogue (arXiv:1609.07369
). The money plot is Fig 4, where the performance of the (less accurate) distances from Gaia is compared to that from a bunch of Cepheids. I wonder whether the relatively poor performance of their model for distances > 2 kpc might not be a function of their chosen prior being actually optimised for the end-of-mission Gaia data, rather than for DR1 data used here.