Masao Sako (Penn) gave the astro seminar today, talking about supernova cosmology, now and in the near future. Afterwards we discussed the possibility that precise cosmological measurements may be reaching their maximum possible precisions, some from cosmic variance and some from complicated and random systematic issues (unknown unknowns, as it were).
Before and at lunch, CampHogg discussed the chapters and title for Hou's PhD thesis, which is about probabilistic inference in the exoplanet domain. This subject of discussion was inspired by Hou's extremely rapid write-up of his new MCMC method (which he is calling multi-canonical, but which we now think is probably a misnomer).
"multi-canonical"
ReplyDeleteI really want to know about these developments! I think anything could justifiably be called multicanonical if it uses many probability distributions of the form (log prior) + c*(log likelihood), which are fairly analogous to the canonical distributions of statistical mechanics.
I still don't really understand how such things could outperform Nested Sampling unless they somehow used more "prior information" about the shape of the likelihood function. We do know that the stretch move doesn't work so well with the hard edges Nested Sampling puts in (at least if you don't marginalise over the "level index"), so that might explain some of it.
We use LOTS of information about the posterior to set up a trial distribution. More soon.
ReplyDelete