PYthon library for Rapid Investigation of (foreground) Components.
The pyric library is meant to allow for users to 'rapidly' do semi-realistic Bayesian component separation in a non-prohibitive time-scale. In practice, this means that currently, the library is limited to
- Diagonal noise matrices,
- Low and common resolutions (nside=16 is the recommended resolution),
- Uniform spectral indices.
The intended flow of analysis is as follows:
- Create a context using the
Contextclass found incontext.py. The context should contain- A list of bands, created using the
Bandclass found inband.py. Each of these represent a detector with a given bandpass and a given set of observations. - A list of components to include in the analysis, using the
Componentclass in thecomponent.pymodule. Currently three kinds of components are possible: CMB components, modified blackbody components, and powerlaw components. An analysis can contain several instances of the same type of component - this means that the sampler will try to fit those components simultaneously.
- A list of bands, created using the
- Once the context has been created, you can run a Gibbs loop using that context by calling
run_gibbs_samplerwithinutils.py.
See driver.py for a working example. In order for this script to work, you must either pass a proper filename to the initialization routines found in component.py so that they match files on your computer, or you must change these filenames in the code itself.
The data currently used to simulate a sky is
- A best-fit powerspectrum file (default is found here:
https://pla.esac.esa.int/pla-sl/data-action?COSMOLOGY.COSMOLOGY_OID=2800) - A dust template, currently the Planck 353 GHz map:
https://pla.esac.esa.int/pla-sl/data-action?MAP.MAP_OID=14613 - A 'synch' template (really just the 30 GHz Cosmoglobe DR1 map):
http://sdc.uio.no/vol/cosmoglobe-data/cosmoglobe_DR1/CG_030_IQU_n0512_v1.fits
Currently, pyric depends on numpy, scipy, astropy, healpy and matplotlib and imageio (for plotting results).