Implement Priors from "Vanilla Bayesian Optimization Performs Great in High Dimensions"

experimental-design / bofire

Experimental design and (multi-objective) bayesian optimization.

https://experimental-design.github.io/bofire/

BSD 3-Clause "New" or "Revised" License

232 stars 23 forks source link

Closed jduerholt closed 5 months ago

jduerholt commented 5 months ago

This PR implements the dimensionality scaled priors from the paper "Vanilla Bayesian Optimization Performs Great in High Dimensions" from Hvarfner et al. (https://arxiv.org/pdf/2402.02229) This is also a nice blogpost regarding the method: https://www.miguelgondu.com/blogposts/2024-03-16/when-does-vanilla-gpr-fail/

Probably, I will set the prior also as new default.

bertiqwerty commented 5 months ago

That's a great idea! I will have a look.

jduerholt commented 5 months ago

Anything missing fore merging it in? Or do you need a bit more time for review?