optuna/optuna-examples

Pruning only on some models?

fickas opened this issue · 2 comments

fickas commented

I do not see how I can define a pruner for something like logistic regression. I do not see what the step value would be. I currently set it to 1 which in essence does no pruning I believe.

Also, am I correct that I can tune KNN by using n_neighbors as the step?

nzw0301 commented

I do not see how I can define a pruner for something like logistic regression. I do not see what the step value would be. I currently set it to 1 which in essence does no pruning I believe.

In an iterative algorithm like gradient descent, step would be iteration or step and value would be validation loss or some metric which should have consistent direction with the objective value.

Also, am I correct that I can tune KNN by using n_neighbors as the step?

It sounds wrong.

nzw0301 commented

Anyway, this question is not related to this repo. Please use https://github.com/optuna/optuna/discussions to ask questions about optuna.