search

mljar / mljar-supervised

Python package for AutoML on Tabular Data with Feature Engineering, Hyper-Parameters Tuning, Explanations and Automatic Documentation
https://mljar.com
MIT License
3.07k stars 411 forks source link

Add Tabnet algo #438

Closed bobcolner closed 1 year ago

bobcolner commented 3 years ago

TabNet: Attentive Interpretable Tabular Learning

Would be great to add this new deep/attention neural net type to mljar. Performance is competitive with boosted trees and should add diversity to ensembles.

https://github.com/dreamquark-ai/tabnet

pplonski commented 3 years ago

@bobcolner have you used TabNet? Does it work good?

Maybe this can be added as the extension to mljar-supervised? (not in the main package, but installed additionally)

pangjac commented 1 year ago

Hi @pplonski Interested in contributing adding TabNet into mljar-supervised, may I get some thoughts on your comments "as the extension .... installed additionally" ? what might be an ideal way to add it "installed additionally" ( in my old mind, I am still stuck in the prototype where adding the functionality -> update dependency -> pip install , so I am not sure what would be "installed additionally") Thanks for the hints and inputs!!

pplonski commented 1 year ago

It will be extra package installed by:

pip install mljar-supervised[tabnet]
pplonski commented 1 year ago

I'm closing for now. I don't want to extend required packages, with tabnet it will require pytorch ...