Séminaire Univers / |
| « A fast stochastic approach for cosmological constraints using weak-lensing peak counts » |
|
Chieh-An Lin |
|
Peak counts from weak lensing (WL) have been proven to be a powerful tool for cosmology. Peaks are direct tracers of the mass function and retain non-Gaussian information, which is complementary to the two-point statistics.
I will introduce a new model to predict WL peak counts, with three main advantages: fast, flexible, and full PDF information. It is fast because the computation time is reduced by many orders of magnitude compared to costly N-body simulations. It is flexible because due to its forward nature, including additional features (e.g. intrinsic alignment) and observational effects (e.g. realistic photo-z) is straightforward. Our model is stochastic so it provides full PDF information of observables, and this makes constraining methods other than the Gaussian likelihood possible. In my talk, I will first describe our model and validate with by comparing with N -body simulations. Then, the impact from the cosmological-dependent covariances will be examined. I will also introduce a robust and efficient constraining method called approximate Bayesian computation (ABC), which gives a parameter posterior without evaluating likelihood. Our study shows a good agreement between ABC and the Gaussian likelihood, while the computation cost for ABC is two orders of magnitude lower. We claim that a fast-stochastic-forward model like ours combining with ABC may have a great potential in future cosmological studies. |
|
mardi 8 septembre 2015 - 11:15 Salle 281 Institut d'Astrophysique de Paris |
| Pages web du séminaire / Seminar's webpage |