Masahiro Nomura

This can be a good candidate: https://github.com/pfnet/pysen

@c-bata Thanks! LGTM!

@c-bata I just published the first version of the CMA-ES with Margin implementation. I’d be happy if i could get your feedback (, though I may not have much time...

@dietmarwo Thanks for interesting information! It is interesting to compare the CMA-ES with Margin with other handling methods including yours. I believe the idea of CMA-ES with Margin can be...

@knshnb Thank you for having interest! `discrete_space=np.array([[0, 10]])` means that the discrete space is {0, 10}. If you want to set the discrite space to {0, 1, 2, ..., 9,...

@knshnb > The current implementation does not seem to support discrete search spaces with different numbers of elements (please see the script below). Is it technically possible to support such...

What do you think? @c-bata

Ok, thanks! Btw, I can work for removing scipy in the next month (if you hope so). When we're implementing it from scratch, I think it is better to compare...

In addition to @c-bata's comment, using CMA-ES with Margin for problems with only binary variables is not appropriate. Actually, Fig.4 in [this paper](https://arxiv.org/abs/2305.00849) depicts that other methods such as (1+1)-CMA-ES...

Thank you for the detailed situation. Maybe I've got the situation. > With Optuna, I've tried the CMA-ES sampler with and without margin. Without margin, it makes essentially no progress....