contramundum53

The problem is that introducing mere possibility for this kind of visualization requires some change in each sampler. As you can see in the above code, I added a logging...

@nzw0301 Yes, by _hook_ I meant `algorithm_logger_callback` argument in the above code.

@blazespinnaker Thanks for your opinion! I think even if we only make visualization public, the implementation would very likely use a hook like the above code. So let me rephrase...

One example I come up with is the following scenario: Currently, TPE estimates the `below` distribution and the `above` distribution separately, and both distributions need to be passed to the...

Another (less aggressive) example would be: Currently `BoTorchSampler` supports `CategoricalDistribution` by transforming categorical variables into one-hot vectors and treating them as continuous variables. In this case, it would be convenient...

Putting experimental labels would be good, I suppose. (Maybe forever experimental?)

@blazespinnaker Sorry for replying late. We had a (rather lengthy) discussion about this issue among the core development members. The conclusions are the followings: * We won't add an API...

@jmsykes83 Sorry for replying late, I had been busy these days. It looks that this PR only changes `target_name`. Does it reduce warning messages?

@toshihikoyanase @not522 Could you check this PR?

@resnant Sorry, `botorch` is in the _optional_ dependency of `optuna` or `optuna-dashboard`. Please install all optional dependencies with ``` pip install optuna[optional] ``` or install dependencies separately as it requires....