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
Is it possible to add good default conservative ranges for each of the hyperparameters for tuning purposes in the documentation? By this I mean something like [0, 100] instead of [0, Infinity] based on results from testing on a lot of different datasets so that these ranges can directly be used for tuning without worrying about what the right values should be each time?
Another alternative could be to embed this info into the algorithm class itself so that HPO packages can directly pick it up
The text was updated successfully, but these errors were encountered:
@trivialfis that issue is about static default parameters. I'm looking for default parameter ranges which can then be tuned using an optimization algorithm like BOHB
Is it possible to add good default conservative ranges for each of the hyperparameters for tuning purposes in the documentation? By this I mean something like [0, 100] instead of [0, Infinity] based on results from testing on a lot of different datasets so that these ranges can directly be used for tuning without worrying about what the right values should be each time?
Another alternative could be to embed this info into the algorithm class itself so that HPO packages can directly pick it up
The text was updated successfully, but these errors were encountered: