What precludes including MLSL and MLSL_LDS for NLOptNoGrad?

[123epsilon]

Just wondering since I was interested in evaluating MLSL and MLSL_LDS as acquisition optimization functions, what makes their implementation different from the algorithms already present in the NLOptNoGrad file? Is it the necessity for a local optimizer?

I'd be willing to implement this but I'm just curious why its not straightforward

[costashatz]

Hello! Thanks for your interest!

I think the honest answer is that they needed a bit of extra work (because of the local optimizer) and we didn't intend to use them, so we did not put this extra effort.

Their implementation should be very much straightforward though. Feel free to propose an implementation. We'll be happy to review/accept it.

[123epsilon]

Awesome, I'll see if I have some time this week to provide an implementation - thanks for the swift response!