Day 917

Thursday.  Friday.


1506.00490
ANNz2 - Photometric redshift and probability density function estimation using machine learning methods
Sadeh, Abdalla Lahav

ANNz2 is a new implementation of the photo-z estimation of Collister and Lahav (2004).  Large photometric galaxy surveys are important for cosmo studies, and in particular for characterizing the nature of DE.  The success of such surveys greatly depends on the ability to measure photo-zs, based on limited spectral data.  ANNz2 utilizes multiple machine learning methods, such as ANN, boosted decision/regression trees and k-nearest neighbors.  The objective of the algorithm is to dynamically optimize the performance of the photo-z estimation, and to properly derive the associated uncertainties.  In addition to single-value solutions, the new code also generates full PDFs in two different ways.  In addition, estimators are incorporated to mitigate possible problems of spec training samples which are not representative or are incomplete.  ANNz2 is also adapted to provide optimized solutions to general classification problems, such as star/galaxy separation.  Illustrate the functionality of the code using data from SDSS DR10 and BOSS.  The code is available publicly.


1506.00665
Galaxy and mass assembly (GAMA): a study of energy, mass, and structure (1kpc-1Mpc) at z<0.3
Driver

GAMA has completed spec-z campaign of >250k galaxies (r<19.8 mag), and will shortly complete the assimilation of the complementary panchromatic imaging data from GALEX, VST, VISTA, WISE, and Herschel.  In the coming years the GAMA fields will be observed by ASKAP allowing a complete study of the stellar, dust, and gas mass constituents of galaxies within the low-z universe (z<0.3).  The science directive is to study the distribution of mass, energy, and structure on kpc-Mpcscales over a 3B yr timeline.  This is being pursued both as an empirical study in its own right as well as providing a benchmark resource against which the outputs from numerical simulations can be compared.  GAMA has 3 particularly compelling aspects which set it apart: completeness, selection, and panchromatic coverage.  The very high redshift completeness (~98%) allows for extremely complete and robust pair and group catalogues; the simple selection (r<19.8mag) minimizes the selection bias and simplifies its management; and the panchromatic coverage, 0.2um-1m, enables studies of the complete energy distributions for individual galaxies, well defined sub-samples, and population assembles (either directly or via stacking techniques).  For further details and data releases, see gama-survey.org.