Improve SkyCube convolution and npred computation #766
Conversation
| energy_axis = LogEnergyAxis(energy, mode='edges') | ||
| elif mode == 'center': | ||
| energy = Energy.equal_log_spacing(emin, emax, enumbins, eunit) | ||
| energy_axis = LogEnergyAxis(energy, mode='center') |
There was a problem hiding this comment.
Please always end with
else:
raise ValueError(...)
for multiple-choice based on user string input.
It really helps to get a good error message when you mistype something.
| 2-dim sky image | ||
| """ | ||
| data = self.data[idx] | ||
| return SkyImage(name=self.name, data=data, wcs=self.wcs) |
There was a problem hiding this comment.
Return copy for data and wcs, at least by default, to avoid accidentally coupled objects?
(I thought we had that method already, didn't review the SkyCube class again. Just ignore me if this is not a good suggestion)
| Parameters | ||
| ---------- | ||
| kernels : list | ||
| List of 2D convolution kernels. |
There was a problem hiding this comment.
What are kernels? Numpy arrays? Astropy kernel objects?
Above you say:
Convolve cube with a given energy dependent PSF.
which suggests the input would be such a PSF object, which I guess it isn't?
Can you make this more clear in the docstring or put an example here or in the high-level docs and link to it from here?
| Returns | ||
| ------- | ||
| kernels : list | ||
| List of 2D image kernels. |
There was a problem hiding this comment.
Say what type of object the kernels are.
| Reference sky cube. | ||
| kwargs : dict | ||
| Keyword arguments passed to | ||
| `EnergyDependentTablePSF.table_psf_in_energy_band()`. |
There was a problem hiding this comment.
Does this resolve as a Sphinx link in the HTML docs?
adonath
force-pushed
the
rework_cube_convolve_npred
branch
from
4f75afd
to
6a1272c
Compare
|
@cdeil Thanks! I've addressed all your comments and will merge as soon as Travis-CI passed. |
adonath
force-pushed
the
rework_cube_convolve_npred
branch
from
b317bfe
to
e7ece72
Compare
cdeil
changed the title
This is the third part of the major
SkyCuberework. These are the most important changes in this PR:convolve_cubefunction toSkyCube.convolve()cube_to_specfunction toSkyCube.to_spectrum, but currently not working.EnergyDependentTablePSF.kernels()method to get a set of kernels for cube convolutioncompute_npred_cubefunctionWhen running the script
docs/tutorials/npred/npred_convolved.py, I get very good agreement with the result from the fermi science tools. The total number of counts agrees within >1%, depending on the precision of the flux image integration.