Is Bayesian Optimization only used for hyperparameter optimization?


With the rise of popular BO libraries like GPyOPT, HyperOPT and of course SigOpt more and more companies and researchers are looking to BO as a way of conducting sample efficient experiments.

For a lot of real world experiments whether it is in fields such as material science, drug discovery or applied physics the price of evaluating a single configuration is extremely high! Therefore, people are looking to BO in order to learn as mush as possible while using the least amount of resources.

If you want to see examples, check out this presentation by Dr. Vishwanath Hegadekatte from the SigOpt Summit where he is discussion BO in the context of material science, check check out this post talking about how BO can be used to produce better glass.

These are just a few examples where BO is used for things other than hyperparameter optimization.