SAMs calibration using Particle Swarm Optimization (ENVIADO AGOSTO 2012)

ANDRES NICOLAS RUIZ; MARIANO JAVIER DE LEON DOMINGUEZ ROMERO; NELSON DAVID PADILLA; SOFIA CORA; DIEGO GARCIA LAMBAS; TOMAS TECCE; IGNACIO GIARGULO; ALEJANDRA MUÑOS ARANCIBIA

MONTHLY NOTICES OF THE ROYAL ASTRONOMICAL SOCIETY

WILEY-BLACKWELL PUBLISHING, INC

Lugar: Londres; Año: 2012

0035-8711

In this work we present an accurate and fast framework designed to select the optimal set of parameters in semi-analytic models (SAM) of galaxy formation. The approach is based on the comparison of the results of the model against a set of observables and the application of improved stochastic techniques in the exploration of the SAM parameter space. Our approach uses the Particle Swarm Optimization (PSO) technique, a self-learning algorithm that surpasses the traditional Markov Chain Monte Carlo (MCMC) sampling methods in terms of computational cost for localizing minima in multi-dimensional spaces. We apply this approach to a particular semi-analytic model called SAG. We choose five parameters from those present in the formulation of the baryonic processes followed within merger trees extracted from a standard CDM N-body simulation and assign a fixed value to the rest. Using PSO we recover the set of best fit SAG parameters and analyze the nearby likelihood surfaces using MCMC in order to explore the intrinsic degeneracies around it. This new methodology is  100 times faster than MCMC methods, and allows us to explore the multi-dimensional parameter space, making fast estimations of the best fit set of parameters following the inclusion of new astrophysical processes involved in the evolution of the galaxy population.