Add observation weights

jolars / sgdnet

Fast Sparse Linear Models for Big Data with SAGA

https://jolars.github.io/sgdnet/

GNU General Public License v3.0

5 stars 2 forks source link

Add observation weights #29

Open wzzlcss opened 5 years ago

wzzlcss commented 5 years ago

Hi mentors, this PR implements observation weights. The changes I am making are multiplying sample weights to gradient, loss and changing to weighted mean in null model, as well as add weight when computing maximum column norm for step size and largest lambda. It seems that glmnet's largest lambda is different.

michaelweylandt commented 5 years ago

Just a quick scan but this all looks reasonable. If we can add some tests and it passes, this should be an easy merge.

wzzlcss commented 5 years ago

Hi mentor, I am finishing poisson and am going to finish this one right away.

michaelweylandt commented 5 years ago

@wzzlcss, is there anything else you wanted to do on this PR or should I go ahead and merge it?