Open AccessFast cosmological parameter estimation using neural networks
Author(s) -
Auld T.,
Bridges M.,
Hobson M. P.,
Gull S. F.
Publication year - 2007
Publication title -
monthly notices of the royal astronomical society: letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.067
H-Index - 122
eISSN - 1745-3933
pISSN - 1745-3925
DOI - 10.1111/j.1745-3933.2006.00276.x
Subject(s) - cosmic microwave background , cmb cold spot , artificial neural network , cosmic variance , algorithm , physics , parameter space , cold dark matter , planck , code (set theory) , cosmic cancer database , astrophysics , spectral line , dark matter , computer science , set (abstract data type) , statistics , mathematics , artificial intelligence , astronomy , quantum mechanics , anisotropy , programming language
We present a method for accelerating the calculation of CMB power spectra,matter power spectra and likelihood functions for use in cosmological parameterestimation. The algorithm, called CosmoNet, is based on training a multilayerperceptron neural network and shares all the advantages of the recentlyreleased Pico algorithm of Fendt & Wandelt, but has several additional benefitsin terms of simplicity, computational speed, memory requirements and ease oftraining. We demonstrate the capabilities of CosmoNet by computing CMB powerspectra over a box in the parameter space of flat \Lambda CDM models containingthe 3\sigma WMAP1 confidence region. We also use CosmoNet to compute the WMAP3likelihood for flat \Lambda CDM models and show that marginalised posteriors onparameters derived are very similar to those obtained using CAMB and the WMAP3code. We find that the average error in the power spectra is typically 2-3% ofcosmic variance, and that CosmoNet is \sim 7 \times 10^4 faster than CAMB (forflat models) and \sim 6 \times 10^6 times faster than the official WMAP3likelihood code. CosmoNet and an interface to CosmoMC are publically availableat www.mrao.cam.ac.uk/software/cosmonet.Comment: 5 pages, 5 figures, minor changes to match version accepted by MNRAS letter