Carrying on with our discussion on Thursday, I generated the “number within 126 pixels vs magnitude of brightest object within 126 pixels” plot. Not as nice a correlation as I think we expected. We can check the code if we want to, but I’m decently confident that I have it right.
To get cutouts of different densities (which the simulation code is now adjusted to run with), I first selected 3 regions that I had a decent number of cutouts for.
Following this, I gathered a bunch of cutouts and visually selected 20 from each. (I can select more or remove certain ones if we want more).
We can look at these in greater depth during the meeting.
As for the Cedar job, it’s still running! This is not surprising to me, as 10x the objects will be roughly 10x the run-time. The place where I divide the code up to multiple cores is on the bin-level, so the run-time will scale with the number of objects. No crashes this time though, so it should run to completion (even if we are waiting a while longer). Fingers crossed! If this isn’t working, I could in theory alter the code to rune in multiprocessing mode on the object level, which will make single bins run faster. Not entirely sure if it is worth the hassle, but it’s possible with a couple hours of troubleshooting.
HSC Research 