UV to k-space pixel maping of output

[piyanatk]

A functionality to map (or match) uv pixels of the output to given k-pixels (e.g. from an already existing image cube) would be very useful for incorporating 21cmSense into other simulations (e.g. generating training sets for machine learning).

[steven-murray]

@piyanatk is the idea that you want Image-space sensitivities? Or just that you want a function to convert uv to k?

[piyanatk]

@steven-murray The idea was just the latter of what you said, but being able to make an image cube of sensitivities would be awesome of course. This idea came from a conversion in the machine learning telecon a while ago that a k-cube or image-space cube would make it convenient for various machine learning purpose.

[steven-murray]

There is already a function in the code that converts uv to k: py21cmsense.conversions.dk_du. As for the latter, see #33 :-)

[piyanatk]

Hm, I am not sure why this conversation was brought up anymore. The function seems to exist since 15 months ago too ...

Anyway, that is great. Just to make sure I understand. The uv grid can be specified in py21cmsense.observation.Observation.ugrid, and because the mapping between Image to K to UV is linear, ones can trace back from the images that they have to to ugrid that they need to specify, correct?

[steven-murray]

Yes, I believe that is all correct :-)