Information-based Mutlifidelity #533
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR introduces an information based approach to multifidelity, using the Multifidelity Maximum Entropy Search method [Takeno 2020, Folch 2023].
This also converts the existing MultiFidelityStrategy to use a separate MultiFidelityAcquisitionFunction. This allows the acquisition function to be explicitly designed by the user. Also, by using BoTorch-style acquisition functions, we are able to generate batches of proposals by conditioning on pending experiments.
Leaving this as a draft because the information-based method doesn't seem to perform very well, which needs some investigation. I'm open to ideas!