Skip to content

Implement AxClient.get_best_parameters, returning the parameter configurations on the Pareto frontier  #656

@josalhor

Description

@josalhor

Hello again!

I've reviewed 22acc17 and every change is excellent! I do however find one thing missing from the API, I cannot find an obvious way to get the points on the Pareto frontier. On the one side, gest_best_parameters is deprecated for multi-objective optimization, returning a random point on the Pareto frontier. On the other, compute_posterior_pareto_frontier requires primary and secondary metrics (the reason for which I am not sure I fully understand).

Am I missing a method/function?

Certainly, one could manually compute the Pareto frontier with get_trials_dataframe, but that doesn't look like an optimal solution.

Edit: For reference, in #654 the example uses compute_posterior_pareto_frontier with only two objectives.

Metadata

Metadata

Labels

enhancementNew feature or requestwishlistLong-term wishlist feature requests

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions