Pico: Parameters for the Impatient Cosmologist

Abstract

We present a fast, accurate, robust and flexible method of accelerating parameter estimation. This algorithm, called Pico, can compute the CMB power spectrum and matter transfer function as well as any computationally expensive likelihoods in a few milliseconds. By removing these bottlenecks from parameter estimation codes, Pico decreases their computational time by 1 or 2 orders of magnitude. Pico has several important properties. First, it is extremely fast and accurate over a large volume of parameter space. Furthermore, its accuracy can continue to be improved by using a larger training set. This method is generalizable to an arbitrary number of cosmological parameters and to any range of l-values in multipole space. Pico is approximately 3000 times faster than CAMB for flat models, and approximately 2000 times faster then the WMAP 3 year likelihood code. In this paper, we demonstrate that using Pico to compute power spectra and likelihoods produces parameter posteriors that are very similar to those using CAMB and the official WMAP3 code, but in only a fraction of the time. Pico and an interface to CosmoMC are made publicly available at http://www.astro.uiuc.edu/~bwandelt/pico/.