Hi there!
I really love LightGBM's speed & performance all into one.
I want the model to output mean and variance, to be optimized using Maximum Likelihood Estimation. I'm guessing I would need to...
> I can give some examples but it will take time to dig up some old notes, need to sort out other things first.
No issues, please take your time.
> In general, it's just taking partial derivative...
> An in-depth example can be found in the master branch. https://github.com/dmlc/xgboost/blob/master/doc/tutorials/advanced_custom_obj.rst
Thanks, but the example shown seems to be a single output...
Is it possible to perform maximum likelihood estimation?
I want the model to output mu and sigma, and minimize the negative log likelihood using a custom loss.