Constructing Cosmological Emulators from a Mixture of Complete and Partial Simulation Results

• Activity Number: 334 - Statistical Methods in Astronomy, Astrophysics and Cosmology
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Astrostatistics Special Interest Group
• Author(s): Earl Christopher Lawrence*
• Companies: Los Alamos National Laboratory
• Keywords: cosmology; emulation; Gaussian process; basis decomposition

Abstract

Cosmology simulations can be incredibly computationally intensive. Thus, there is great interest in statistical emulators, like the Cosmic Emu, that give high fidelity predictions for features of these simulations, but at much faster speeds. Because of the computational expense required to generate a large training set of complete high resolution simulations for the emulator, we supplement the training set with various types of low resolution results. This talk will discuss how to combine high resolution results for the matter power spectrum with low resolution results that only return a portion of this function. Our approach exploits the basis decomposition for emulation described in Higdon et al. (2008). We use the partial results to produce a "complete" set of data in weight space and use this to build the emulator.