contramundum53

Thanks for your suggestion! In that case, what about using a log scale all the way to the end? I think switching to linear scale doesn't help much if, e.g.,...

This is possible by specifying [StaticErrorEvaluator](https://github.com/optuna/optuna/blob/master/optuna/terminator/erroreval.py#L109).

Thanks for your suggestion! I'm personally interested in these kind of techniques, especially since there is a trend training very large models these days. I'll have a look at the...

Thank you for creating the issue. > Now, out of curiosity, what is the difficulty with adding a study attribute to FixedTrial? My personal opinion is that 1. _requiring_ a...

But indeed it might be nice to have an official optional `study` argument. Let us consider this again when there are more people who want this feature.

By the way, even if we add `study` attribute some day, I personally don't think we will be breaking the code ``` fixed_trial.study = some_study ```