Lognormal sampling#73
Conversation
…owing a full covariance with a simple basic approximation
…, as well as a sims_general module to try something enw
… map (convergence map)
…lter to create a mask for point sources
…tions in poisson map, simple match filter to create a mask for point sources
Sebastian-Belkner
left a comment
Sebastian-Belkner
left a comment
There was a problem hiding this comment.
This PR introduces two great new features to delensalot.
-
The first is a new iterator that finds the MAP point by assuming that phi is lognormally distributed. It also provides ways to generate lognormally distributed simulations that can be used for the reconstruction. This is in particular interesting for delensalot as we know that phi is not perfectly Gaussian. While the standard delensalot iterators work with non-gaussian phis by assuming it to be Gaussian, a reconstruction that supports non-gaussian phis is a great extension.
-
The second is an extension of the simulation library. The PR introduces the generation of point source foregrounds and is embedded in the already existing simulation library.
I am happy to approve it once the comments and merge conflict are resolved.
| tlm = self.unlcmbs.get_sim_tlm(self.offset_index(idx, self.offset_cmb[0], self.offset_cmb[1])) | ||
| dlm = self.get_sim_plm(idx) | ||
|
|
||
| hp.write_alm(pfname, dlm) |
There was a problem hiding this comment.
This could lead to unexpected behaviour. A function returning tlms shouldn't write dlms.
|
|
||
| if (i + 1) in itrs: | ||
| ret.append(rlm2alm(rlm)) | ||
| print("Doing for kappa0", kappa0) |
There was a problem hiding this comment.
This looks like a huge print for each iteration. Suggest to use log.debug() instead
|
|
||
| @staticmethod | ||
| def load_plms(lib_dir, itrs): | ||
| def load_plms(lib_dir, itrs, kappa0 = None): |
There was a problem hiding this comment.
What does the kappa0 do? Docstrings would help
| rlm2alm = lambda x : x.copy() | ||
|
|
||
|
|
||
| from plancklens import shts |
There was a problem hiding this comment.
plancklens.sht only part of master branch, delensalot however uses plancklens/plancklensdev. This line will therefore fail.
Suggesting to use lenspyx shts directly, as it is done throughout delensalot. Also, plancklens dependecy will be removed in the near future.
|
|
||
| from . import cs_iterator as csit | ||
|
|
||
| alm2rlm = lambda alm : alm # get rid of this |
There was a problem hiding this comment.
I think they can be deleted
|
|
||
| from plancklens import shts | ||
| from delensalot.utility.utils_hp import Alm, almxfl | ||
| from plancklens.utils import cli |
There was a problem hiding this comment.
delensalot has its own utils.cli()
| log.info("\r [" + ('%02d:%02d:%02d' % (dh, dm, ds)) + "] " + label) | ||
| return | ||
|
|
||
| typs = ['T', 'QU', 'TQU'] |
There was a problem hiding this comment.
I don't think this is used
|
|
||
| def transform(rlm, kappa0): | ||
| if kappa0 is not None: | ||
| print('kappa0 is not None, transforming to phi') |
There was a problem hiding this comment.
Suggesting to use log.warning or log.debug instead, so there is context information and so that it can be controlled via verbosity
There was a problem hiding this comment.
This implementation uses the pertmf iterator. Does this also work for the const_mf iterator?
The procedure of your implementation will be described in the n32 paper, but having the outline described in the docstring would be very helpful for future reference
|
|
||
|
|
||
| class iterator_pertmf(csit.iterator_pertmf): | ||
| def __init__(self, lib_dir:str, h:str, lm_max_dlm:tuple, |
There was a problem hiding this comment.
There are a few new parameters that lack docstrings
|
@Saladino93 what's the status here? looks like we have merge conflicts |
2 participants
Here I implement a version where I sample over the Gaussian field of a log-normal field. Which tests should I do?
The relevant files are:
cs_iterator_lognormal.py and the stepper.py
All the others can be ignored (foreground, iterative bh, etc...) as they are part of other branches.