For seminars and other local announcements, please subscribe to sfu-cosmo mailing list.
Tuesday, 10 March 2009, 14:00 in P8445A
Prof. Phil Gregory (UBC)
Mining for extra-solar planets with a Bayesian MCMC Kepler periodogram
Are we on the threshold of discovering life elsewhere in the universe? With the mounting discoveries of extra-solar planets there is a sense that we are well underway. Astronomers searching for the small signals induced by planets inevitably face significant statistical challenges. This talk illustrates how a Bayesian re-analysis of published radial velocity data sets is providing strong evidence for additional planetary candidates. The Bayesian nonlinear model fitting is accomplished with a new hybrid Markov chain Monte Carlo (HMCMC) algorithm which incorporates parallel tempering, simulated annealing and the genetic algorithm. Each of these components facilitate the detection of a global minimum in χ2. By combining all three, the HMCMC greatly increase the probability of realizing this goal in a blind search of the parameter space. When applied to the Kepler problem it acts as a powerful multi-planet Kepler periodogram for both parameter estimation and model selection. The algorithm is embedded in a unique two stage adaptive control system that automates the tuning of the MCMC proposal distribution parameters through an annealing operation. The HMCMC algorithm is very general and applicable to many other nonlinear model fitting problems.
Seminars in 2008:
[ See complete seminar archives | iCal feed ]
Modified by Andrei Frolov <frolov@sfu.ca> on 2023-11-01