Starting AceNet Processing
We are fully into production mode now! Out of the 10 tracts that exist in our sample, I have finished 4 of them and am currently fitting 9812 and 9814. The largest tract, 9813, is something I am going to try and tackle tomorrow. I plan to have all the processing done and my results moved off the cluster before Cedar goes into a 48 hour shutdown in 4 days.
One tract, 9571, returned 48320 galaxies! Assuming an even distribution across 5 bands (u and u* images only exist for 9813 and 10054), that’s over 9646 unique galaxies! Wow! I will also note this is for one of the smaller tracts. We should expect MUCH more for 9813 and 9812.
Finished F2 Field Processing
All possible profiles for galaxies and stars for the SHELS F2 field have been extracted. I wrote a utility script that compiles together 1 FITS file containing every unique object, so we don’t have a large set of individual files that we need to sort through. This will also be helpful once all our CLAUDS-HSC data is obtained.
I am currently working on a Jupyter notebook script that generates PSFs for our galaxies. This will save a third file, F2_PSFS.fits, which will have PSFs that can be ID matched. Configuring my Sersic fitting code to work with this data should be very simple!
Ivana, do you happen to have the data for the SHELS F1 field? I was thinking it might be worth it to also extract profiles for that field as well. Seeing as on AceNet, fitting the 20000 or so objects in F2 (duplicates included) took about 90 minutes, it wouldn’t be a large burden to gather those profiles to have in tandem with F2.
Pushing even further, the pipeline is now extremely easy to modify for any dataset you have, as it was rewritten with that in mind. Essentially, if you want me to extract profiles for basically anything that is available to you, I am more than willing to do so!
Paper Progress and an Interesting Find
I have also been working on filling out the MNRAS paper on the pipeline. While working last night, I figured it would be a good idea to work out the time per object, and that time whether it is succeeding or failing. I discovered after doing some tests that the time per failure and time per success differs per band! We do typically see a trend (except in the r band) of successful fits taking longer, and mean times hovering around 1 second per object. The reason why z, u and u* are not included is because the linear equations to solve for t_s and t_f didn’t return a valid answer (negative times for either success or failure).
The link to the paper is here, just in case it’s needed! https://www.overleaf.com/read/xsdsfgxprzbv
HSC Research 