Bayesian inference for radio observations …

I noticed this paper on astro ph today by Perkins et al. concerning the MeqTrees code which reminded me that I haven’t yet blogged about the original BIRO (Bayesian Inference for Radio Observations) paper by Lochner et al. from last week.  The goal of the paper is to perform Bayesian inference of flux densities for point sources in a radio image under uncertain calibration: i.e., to marginalise over pointing (and other) errors; and in this specific goal the algorithm proposed is very reasonable, and appears to have been well implemented (cf. the neat posterior covariance matrix illustration in their Fig 6).  Where the authors’ description of BIRO over-reaches is in what I perceive as their attempt to link BIRO with the more general class of algorithms applying Bayesian non-parametrics (esp. GPs) to astronomical imaging problems; e.g. their repeated references to “the image construction problem” with the truth being just “one realisation from an infinite ensemble of images compatible with the data” (and direct citations to Ensslin et al., Junklewitz et al.).  My point being that in fact BIRO is a strictly parametric Bayesian solution and indeed one with (at present) a fairly limited range: the modelling of point sources (with a limited ‘extended source’ alternative) not unlike the much aligned CLEAN algorithm.  If BIRO is to be developed into a more general image reconstruction tool within the parametric realm then the treatment of model selection will have to become much more sophisticated (e.g. ditching the uniform priors [perhaps the ‘worst’ priors for Bayes factor purposes], and introducing a sampler over model space like RJMCMC / birth-death moves).

This entry was posted in Uncategorized. Bookmark the permalink.

One Response to Bayesian inference for radio observations …

  1. Pingback: P-values from posteriors … | Another Astrostatistics Blog

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s