You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There's really no point to optimizing the number of estimators in ensemble models: Basically, the more estimators they're provided, the better they perform. Let's remove n_estimators as an optimizable parameter for all ensemble models and instead default it to, say, 100 or 500 estimators.
The text was updated successfully, but these errors were encountered:
The RF operator has had its n_estimators parameter removed, and it now defaults to 500 for n_estimators. The remaining ensemble operators still need to be reworked.
Increasing the estimators will increase performance, but I would think when you're trying a lot of different parameters it will increase the time for all those decision trees to run, and whether it wouldn't be better to increase n_estimators later. I guess if they don't perform well enough in the first instance they're not going to make it into the next generation. Really just curious on your thoughts.
I suppose the default could be n_estimators=100 and it should still work fine. I just want to make sure that RFs and other ensemble methods don't get skipped over simply because they didn't have enough estimators.
There's really no point to optimizing the number of estimators in ensemble models: Basically, the more estimators they're provided, the better they perform. Let's remove
n_estimators
as an optimizable parameter for all ensemble models and instead default it to, say, 100 or 500 estimators.The text was updated successfully, but these errors were encountered: