Group Leaders: Jeff Jewell (Astro, JPL), Joe Guinness (Stat, NCSU)
SAMSI Webmaster: Hyungsuk Tak
Weekly Meeting: SAMSI Room 203 / Tuesdays 4:00-6:00pm ET
Description: Statistical and computational innovations are urgently needed for analyzing the next generation of cosmological data. This working group will bring together leading researchers in cosmology, computational spatial and Bayesian statistics, experimental design, and computer modeling to develop methodological advances necessary for answering fundamental questions about the origin and large scale structure of the universe.
Big Questions (some topics treated jointly with UQ group):
- How do we perform Bayesian inference for spatial Gaussian random fields in the presence of galactic foregrounds (source separation)?
- How do we design MCMC (or other, such as Hamiltonian sampling) sampling methods for high-dimensional data?
- Bayesian approach to Large-Scale Structure – how to make inferences from redshift survey data?
- How can we make inferences for deterministic nonlinear dynamical systems (inference of initial conditions and model parameters)?
- How do we design emulators for systems with stochastic ICs/BCs and deterministic evolution (e.g., N-body problems)?
- How do we design experiments for doing Bayesian inference for model parameters for problems which involve forward simulation which is very expensive?
- Can we implement fast direct Gaussian process computational methods for analyzing cosmological datasets?
News and Updates: Coming Soon…
SAMSI Directorate Liaison: Sujit Ghosh