clmm Code

Structure

Initial clmm proposed code structure -- outdated!

clmm is a general code for performing individual- and population-level inference on galaxy cluster weak lensing data. It will serve to enable the CLMassMod Key Task of the LSST-DESC SRM and will be used as a framework for future CL WG activities. clmm aims to be modular in (at least) three respects:

1. clmm will be able to run on real data as well as simulations, and it will not be restricted to any particular datasets.
2. clmm will support multiple modes of inference of the cluster mass function and other relevant distributions, such as the mass-concentration relation.
3. clmm will enable evaluation of results on the basis of a number of different metrics, some of which will not require a notion of truth from a simulation.

To enable these goals, clmm will have (at least) three major superclasses of objects:

0. clmm.utils

1. clmm.Reader objects that are able to load data and corresponding metadata from external files; subclass instances will be specific to the native format of the pipeline that produced the data and/or the type of data (shear maps vs. mass surface density maps, etc.) A. config file readers for clmm.Halo.map, clmm.Halo.mass, etc. reading in data itself in different original formats from simulations/observations

2. clmm.Halo class Z. clmm.Halo.config_params A. clmm.from clmm.Halo.Map superclass (???) i. clmm.Halo.map.data['shear'] and clmm.Halo.map.data['convergence'] ii. clmm.Halo.map.metadata including radial range, binning, miscentering, noise level for a profile fit, e.g. iii. clmm.Halo.map.convert('mass2D', 'shear', config_params) (make new maps i.e. clmm.Halo.map.data['shear'] from existing maps i.e. clmm.Halo.map.data['mass2D']) B. clmm.Halo.redshift C. clmm.Halo.mass i. clmm.Halo.mass.MLE ii. clmm.Halo.mass.true_3D D. clmm.Halo.utils i. clmm.Halo.utils.cross_correlator() taking two maps and producing another map E. clmm.Halo.build_profile(clmm.Halo.map) requiring clmm.Halo.config_params and checking if unspecified, producing clmm.Halo.obs_profile i. clmm.Halo.obs_profile as an instance of the clmm.Halo.Profile class

3. clmm.Halo_Model superclass, must have .evaluate_func() (like the NFW functional form) and .evaluate_prior() (such as uniform distribution or Gaussian, etc.) A. consider clmm.NFW subclass taking config file with prior params i. clmm.NFW.evaluate_function(x, model_params) where data can be clmm.Halo.profile and params is list(?), evaluates a function ii. clmm.NFW.evaluate_likelihood((xobs, yobs), model_params) which calls clmm.NFW.evaluate_function() ii. clmm.NFW.evaluate_prior(model_params) evaluates p(model_params | prior_config_params) B. mass aperture has different data, etc.

4. clmm.Inferrer objects that accept data produced by the clmm.Reader subclass object as well as other information, such as theoretical models and selection/systematics functions, and output estimates of the desired quantity, such as a probability distribution over the mass-concentration relation or a stacked estimator of the mass function. A. clmm.Halo_Inferrer(clmm.Halo, config_stuff, **params) class i. produces output in the form of posteriors (or point estimates) ii. clmm.Halo_Inferrer.profile_fitter(clmm.Halo.map, profile_config_params) adding fit info to clmm.Halo.profile, but maybe later because not well defined right now iii. clmm.Halo_Inferrer(clmm.Halo, clmm.Model, config_params) outputting posterior (or samples or MLE) for one halo given config_params for inferrer

5. clmm.Metrics objects that take as input one or more estimators produced by clmm.Inferrer objects and, when available, the true value of the estimated quantity read in by a clmm.Reader object separately.

6. extensions for many-halo inference punting to later A. clmm.Halo_Ensemble i. clmm.Halo_Ensemble.select() that bins clmm.Halo objects by some attribute they have B. clmm.Halo_Ensemble_Inferrer(clmm.Halo_Ensemble, clmm.)

Users and developers may implement their own subclass instances of these superclasses to flesh out the clmm functionality.