Organised by:
in collaboration with:

secretariat:

 

social media:


 



Accesso Utenti

Geneva Observatory, Geneva
Address:
Switzerland
Miscellaneous Information:

Abstract Reference: 30230
Identifier: P1.30
Presentation: Poster presentation
Key Theme: 1 Reduction and Analysis Algorithms for Large Databases and Vice-versa

MCMC algorithms at the service of exo-planets hunters

Authors:
Sosnowska Danuta, Segransan Damien, Diaz Rodrigo, Buchschacher Nicolas, Alesina Fabien

Exo-planetary research is a fast growing science domain. More and more astronomical instruments are dedicated to exo-planet searches, collecting terabytes of data. In order to explore this huge amount of data sophisticated algorithms are necessary. The Markov Chain Monte Carlo is a family of algorithms very well suited for the exploration of the high dimensional space of parameters describing keplerian orbits. The Data & Analysis Center for Exoplanets (DACE) contains a database with thousands of radial velocities measurements and transit light curves amongst other observations. It also implements algorithms for treating, displaying, optimising and exploring this data. One of the recently developed algorithms for fitting keplerian orbits is the MCMC algorithm with two variations: the simple Metropolis-Hastings MCMC and its Change of Basis (CoB) version. The simple MHMCMC is very efficient for well constrained orbital parameters, the CoB works better for less constrained cases, such as when the orbits are not closed. The software was written in Java, whose object oriented structure allows for nice integration of several solutions into the same scheme and separation of the algorithm from the model. In this way, the DACE MCMC can be applied to any astrophysical model within DACE. The DACE MCMC is directly linked to the DACE database, so it can be launched on the data extracted from the DB on the DACE server or locally on the user computer. It can also be run on user private data. The software development is still ongoing and future features will include the incorporation of different models as well as the running of several Markov Chains in parallel with their solutions combined.