Nested optimization of acquisition function

[HanyangHenry-Wang]

In Botorch, Knowledge Gradient is calculated by One-Shot method and I can understand the code. But usually, KG is calculated by a nested optimization problem. My question is that now I would like to build a KG which uses nested optimization, is there any way to do so in Botorch? Or is there any acquisition function in Botorch that use nested optimization? Thank you.

[sdaulton]

I actually just implemented KG with nested optimization. I’ll put up a PR, hopefully in the next week

[HanyangHenry-Wang]

That will be very helpful, thank you!