search

catboost / catboost

A fast, scalable, high performance Gradient Boosting on Decision Trees library, used for ranking, classification, regression and other machine learning tasks for Python, R, Java, C++. Supports computation on CPU and GPU.
https://catboost.ai
Apache License 2.0
7.98k stars 1.18k forks source link

[Question] Target distributions for CatBoostRegressor (with RMSEWithUncertainty loss function) #1571

Open jarandaf opened 3 years ago

jarandaf commented 3 years ago

According to the docs, CatBoostRegressor supports probabilistic forecasts via RMSEWithUncertainty loss function. It seems the implementation is based on NGBoost. Are other target distributions p(y|x) (besides the Normal) supported? The paper does not seem to constrain the learning framework to use a specific probability distribution.

alexandrehsd commented 1 year ago

I'm also interested in this discussion. Also, there is a project aiming for that called CatBoostLSS, but apparently it is a dead project.

Evgueni-Petrov-aka-espetrov commented 1 year ago

hi guys! it is very nice to see that you are looking at catboost library!

@alexandrehsd huge thanks for your comments here!

@jarandaf at the moment, the normal distribution is hard-coded, but we would like to hear about use-cases where other distributions are useful!