search

basf / mamba-tabular

Mambular is a Python package that brings the power of Mamba architectures to tabular data, offering a suite of deep learning models for regression, classification, and distributional regression tasks. This includes models like Mambular, FT-Transformer, TabTransformer and tabular ResNets.
https://mambular.readthedocs.io
MIT License
128 stars 6 forks source link

quantile regression #119

Closed GidonKR closed 1 month ago

GidonKR commented 2 months ago

I would like to use this model to run a quantile regression i couldn't find an example, also i see that you have a full distribution prediction, but what if i don't want to assume any distribution?

Describe the solution you'd like if there is a possibility to run a quantile regression without any distribution assumptions Thanks

AnFreTh commented 2 months ago

hi @GidonKR thanks for reaching out. So far, mambular does not support quantile regression, but if there is real interest, we will include it in the future. Just to clarify: You are looking for vanilla quantile regression, (e.g. minimization of the pinball loss) to get quantile predictions?

GidonKR commented 2 months ago

Hey, thanks for the quick reply, yes i was referring to minimization of the pinball loss I would love to see it in order to compare it with the results i get with catboost.

AnFreTh commented 2 months ago

Alright. I will leave this issue open and close it, once the functionality is implemented.

AnFreTh commented 1 month ago

Included quantile regression with PR#126 and release v0.2.3. I will close this issue for now. Feel free to reopen it or open another issue for further enhancements, or if you have any issues with the current Quantile implementation.