search

drphilmarshall / Pangloss

Line of sight mass reconstruction in the Universe
GNU General Public License v2.0
10 stars 10 forks source link

Notebook index ordering, likelihood demo #99

Open drphilmarshall opened 7 years ago

drphilmarshall commented 7 years ago

Hi @sweverett ! Meet David Thomas ( @davidthomas5412 ), a Stanford masters student who will be working on Pangloss this quarter. We're going to try some simple sampling and optimization experiments, building on your thesis work. To get started, David is working through your notebooks, and adding more documentation. I made an index notebook, with an ordered list of demos he should work through, here. Can you please edit this to get the ordering correct (think: how would you introduce someone to the Pangloss code, step by step?) and to add one line explanations of what is in each notebook? Thanks :-)

Also: I looked for and couldn't see a notebook showing how to calculate the log likelihood of a foreground halo catalog given a background catalog of galaxy shapes. Can you please either rename one or make one? It doesn't have to be very long :-) Thanks!

sweverett commented 7 years ago

Hi @davidthomas5412 ! Glad to have you working on Pangloss with us! (And sorry for the late response - I initially didn't see this issue in the flood of new Pangloss emails :) )

As I was catching up to speed, I unfortunately realized there were a few commits on the bin_to_maps branch that never got pushed once my thesis was finished. The changes were mostly on finishing up the log-likelihood calculation and speeding up the smooth-component correction by pre-calculating the redshift bin volumes in grid.py, so I don't think it will affect the plotting bugs you have been working on. Sorry about that! I'll look into any conflicts with merging the updates into wl.

Speaking of plotting, the notebooks started to fail after an astropy update in between the end of SULI work last summer and when I began work on my thesis. I thought I had fixed most of the plotting bugs but am not surprised you're running into issues. As you pointed out in #101 , there is certainly a better OO approach to the plotting than what we have now. Let me know if you have any questions about the plotting code as you work on improving it!

I'll take a look through the existing demos and write a new one for the likelihood calculation, as well as add some comments on some scripts I used to generate most of the results in my thesis in case they are useful.

drphilmarshall commented 7 years ago

Hi Spencer! Thanks for the commits (being merged in to wl in #104 ) - and for the likelihood demo, too. That'll be most useful for making sure we are on the right track. Much appreciated!

The astropy failures: we're wondering about switching to pandas in any case, so we may postpone making all fixes until that point. Interested to hear any opinions you have on that.