Constraining Cosmological Parameters, Including Neutrino Mass, Using N-body Large Scale Simulations and Artificial Neural Networks

Details Constraining Cosmological Parameters, Including Neutrino Mass, Using N-body Large Scale Simulations and Artificial Neural Networks Constraining Cosmological Parameters, Including Neutrino Mass, Using N-body Large Scale Simulations and Artificial Neural Networks

Jan 2012

adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2012aas...21914309a&link_type=abstract

American Astronomical Society, AAS Meeting #219, #143.09

Statistics

Computation

Scientific paper

Future or ongoing galaxy redshift surveys such as BOSS and EUCLID will explore the nature of dark energy using the expansion history of our Universe and clustering information of large-scale structure. These surveys promise to achieve high-precision measurements of galaxy power spectrum amplitudes to 1%-level precision and offer a possibility to improve constraints on cosmological parameters, including neutrino masses. The standard linear theory of structure formation can not be used to make theoretical predictions on scales smaller than 100 Mpc/h, below which the non-linear effects become significant compared to the precision of future surveys. One approach is to analytically model non-linear matter power spectrum based on higher-order perturbation theory. However, at redshift z=0, the perturbation approach is expected to reproduce the N-Body results within 1% - only for modes with k < 0.1 h/Mpc. The other approach to evaluate the small scale clustering is to run N-Body simulations over a finely spaced grid in multi (of order 10) dimensional parameter space. If each parameter is sampled 10-20 times over its range, it would require a total of 10(10-20) simulations. This is not feasible with each simulation consuming 1000-2000 cpu hours.
Using Artificial Neural Networks (ANN) to model the non-linear matter power spectrum, we show that an optimally trained ANN is capable of reproducing the matter power spectrum obtained directly from N-body simulations, to within 1% precision upto k < 1.0 h/Mpc. Our preliminary analysis has shown that training an ANN requires a suite of at least 200 high-resolution N-Body simulations, which is far more feasible than 10(10-20) simulations. We are currently running these simulations, using the ENZO code developed by the Laboratory for Computational Astrophysics at the University of California in San Diego (http://lca.ucsd.edu). This work is supported by the National Science Foundation through TeraGrid resources provided by the NCSA.

Affiliated with

Also associated with

No associations

Rating

Constraining Cosmological Parameters, Including Neutrino Mass, Using N-body Large Scale Simulations and Artificial Neural Networks does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.

If you have personal experience with Constraining Cosmological Parameters, Including Neutrino Mass, Using N-body Large Scale Simulations and Artificial Neural Networks, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Constraining Cosmological Parameters, Including Neutrino Mass, Using N-body Large Scale Simulations and Artificial Neural Networks will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFWR-SCP-O-1575910