Iterative learning

[gentine]

Would there be a simple way in NGBoost to do some iterative training of a previously trained model (ideally with a different learning rate and on a different dataset)?
Thanks a lot

[alejandroschuler]

I don't think there's a simple way to do this at the moment. To be clear, you're talking about adding base estimators to an already fit model, right? Something like:

model = NGBoostRegressor(n_estimators=100, learning_rate=0.1)
model.fit(X1, Y1)
model.add_estimators(
    X2, Y2, 
    params=dict(n_estimators=50, learning_rate = 0.01)
)

I brought this up briefly in another recent discussion (#253). If you're interested in making a PR, I don't think it would be terribly challenging and it would be much appreciated!

[gentine]

Hi Alejandro, Yes this is exactly what I meant. I will try to to make a PR.

…

On May 19, 2021, at 4:58 PM, Alejandro Schuler ***@***.***> wrote: I don't think there's a simple way to do this at the moment. To be clear, you're talking about adding base estimators to an already fit model, right? Something like: model = NGBoostRegressor(n_estimators=100, learning_rate=0.1) model.fit(X1, Y1) model.add_estimators( X2, Y2, params=dict(n_estimators=50, learning_rate = 0.01) ) I brought this up briefly in another recent discussion (#253 <#253>). If you're interested in making a PR, I don't think it would be terribly challenging and it would be much appreciated! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#256 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADC6VIXRL6V6LVZ3GRDBQXLTOQRAJANCNFSM45FNDJYQ>.