search

microsoft / FLAML

A fast library for AutoML and tuning. Join our Discord: https://discord.gg/Cppx2vSPVP.
https://microsoft.github.io/FLAML/
MIT License
3.85k stars 506 forks source link

Quantile Regression #780

Open velezbeltran opened 1 year ago

velezbeltran commented 1 year ago

Hello!

I was wondering if there is any native way of doing quantile regression with Flaml or any quick way of going about setting it up so that it works with it. I was taking a look at the tutorials and it wasn't clear to me how to do so.

Alternatively, if there isn't one, happy to make a PR implementing it if people support this feature.

Thank you!

sonichi commented 1 year ago

It seems that LGBMRegressor already supports quantile regression. I haven't checked other libraries yet. In LGBMRegressor, we need to add alpha, objective="quantile". To pass these parameters to LGBMRegressor in flaml, the simplest way is to use the "custom_hp" argument in AutoML.fit(). See https://microsoft.github.io/FLAML/docs/Use-Cases/Task-Oriented-AutoML/#a-shortcut-to-override-the-search-space for an example.

velezbeltran commented 1 year ago

Thanks @sonichi will try it out.