search

mclaughlin6464 / pearce

A (better structured) package that implements the calculation of mock cosmological observables and their emulation.
MIT License
0 stars 1 forks source link

New Assembly Bias Formalism #5

Open mclaughlin6464 opened 7 years ago

mclaughlin6464 commented 7 years ago

I've designed a new assembly bias formalism that is a continuous modification of Hearin's decorated HOD. I need to implement in pearce, and then test that it can recreate existing galaxy catalogs which are known to include assembly bias.

mclaughlin6464 commented 7 years ago

This is implemented and for the most part works well. However, there is a bug with the f_c parameter. I need to make it so it is set as the new upper bound for the central object, else you can have object population above that amount.

Come to think of it. I'm not sure if that a problem if the mean is still the same. I'll have to ask someone with a good interpretation of the the true meaning of the f_c parmeter.

mclaughlin6464 commented 7 years ago

Additionally, my formalism works well be seems slow. I should diagnose which portion is slow and see what speedups I can do.

mclaughlin6464 commented 7 years ago

W.r.t. the f_c debate. Joe says that he thinks that it is fine. However, it would be easy to add it into the RedMagicCens object as upper_assembis_bound.

mclaughlin6464 commented 7 years ago

My new extension is a little slow, and i'm working to diagnose why. One thing that sucks is that the assembias for satellites has to repeat the assembiasing of the centrals that it just already did. would be nice if that could be avoided.

mclaughlin6464 commented 7 years ago

I've looked into the slowness a bit. It appears that all the conditional operations are the slowest. I'm going to look into what portions of those are so slow. What is interesting to me is that those operations are very slow for centrals than satellites, and I have no idea why. It's essentially the same calculation on the same halos!

Additionally, it also seems like I get different times for when I'm populating from when I do the calculation directly. I don't know why that is either...

mclaughlin6464 commented 7 years ago

I think I want to add the notion of a " maximum" slope. I've tinkered with this idea before. I think this should be the last think I try to work with.

mclaughlin6464 commented 6 years ago

Haven't updated this in awhile. I've done a lot of work understanding these assembly bias models, and it appears that in general, they are insufficient to jointly describe the HOD and clustering of many occupation distributions. I will leave this open, as it is the subject of ongoing work, but I don't think I'll be adding too many new models to pearce for the time being.