[vervez]

There's also recent Bayesian Optimal Experimental Design methods that allow you to directly design experiments using gradient ascent. Not sure how it compares with BayesOpt on your problem, though.

[capableweb]

That sounds really interesting, do you have more information about it?