||The ChiVO Library: Advanced Computational Methods for Astronomy
||512, Astronomical Data Analysis Software and Systems XXV
||Solar, M.; Araya, M.; Mardones, D.; Arevalo, L.; Mendoza, M.; Valenzuela, C.; Hochfärber, T.; Villanueva, M.; Jara, M.; Simonsen, A.
||The main objective of the Advanced Computational Astronomy Library (ACALib) is
to collect in a coherent software package the research on computational
methods for astronomy performed by the first phase of the Chilean Virtual
Observatory between years 2013 and 2015. During this period, researchers and
students developed functional prototypes, implementing state of the art
computational methods and proposing new algorithms and techniques. This
research was mainly focused on spectroscopic data cubes, as they strongly
require computational methods to reduce, visualize and infer astrophysical
quantities from them, and because most of the techniques are directly applicable
either to images or to spectra.
The ACALib philosophy is to use a persistent workspace abstraction where
spectroscopic data cubes can be loaded from files, created from other cubes or
artificially generated from astrophysical parameters. Then, computational
methods can be applied to them, resulting in new data cube instances or new data
tables in the workspace. The idea is to provide not only API bindings for the
workspace and the cubes, but also web-services to use the library in cloud-based
frameworks and in the Virtual Observatory.
In a nutshell, ACALib is integrating and testing several cube manipulation
routines, stacking procedures, structure detection algorithms, spectral line
association techniques and a synthetic data cube generation module. The library
is developed in python, strongly rooted in astropy modules and using efficient
numerical libraries like numpy and scipy, and machine learning libraries such as
scikit-learn and astroML.
In the near future, we plan to propose ACALib as an astropy affiliated package,
and to create a CASA add-on to ease the usage of our methods. Also, we are
exploring bleeding-edge computational methods to include in ACALib, such as deep
learning networks, and developing new prototypes for other types of astronomical
data, such as light curves in the time-domain.