search

LSSTDESC / rail_base

Base classes for RAIL
MIT License
0 stars 1 forks source link

Weights from working groups in n(z) estimation #3

Open joezuntz opened 4 years ago

joezuntz commented 4 years ago

When the WGs pass you a collection of galaxies we've selected that we'd like the n(z) for, we may also need to pass you a weight per galaxy. The perfect n(z) would then be a weighted histogram of the true z values for each object.

I just wanted to make sure that was on your radar!

aimalz commented 4 years ago

Thanks for bringing this up @joezuntz! The Estimator class can accept subclass-specific information beyond the training and test set files already through its config file, so that functionality will already be built in when it's ready for extension into a Summarizer class. (@danielgruen and @jmccull might be interested in implementing an Estimator subclass that makes use of this functionality and/or writing unit tests for it, which can be started now by branching off this as-yet-unmerged but mature branch.)

eacharles commented 2 years ago

Nzdir seems to have weights that meet the stringent criteria of Joe. The others don't. We should come back to this.