Blackbox function optimization with bounds?

[NeutralKaon]

Hi there! Thanks for making this – it is incredibly impressive.

Some of the optimisers present natively support bounds (e.g. humpday/optimizers/freelunchcube.py contains on line 22 bounds = np.array( [ np.array([0, 1])] * n_dim )). My problem is bounded and rapidly becomes flat / infinite outside of a feasible region.

Is it possible, feasible, or even a remotely good idea to naïvely ask about the possibility of including lb and ub in recommend()?

[microprediction]

Sorry for delayed response. In this benchmark the bounds are always the hypercube. So rather than lb, ub you need to transform the objective