feat: balanced clustering #2999
Draft
Conversation
This file contains 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
This comment was marked as resolved.
This comment was marked as resolved.
jacketsj
force-pushed
the
jack/balanced-clustering
branch
from
b4d8738
to
9493865
Compare
jacketsj
force-pushed
the
jack/balanced-clustering
branch
from
9493865
to
83a30a4
Compare
# for free
to join this conversation on GitHub.
Already have an account?
# to comment
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Outdated suggestions cannot be applied.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 implements FCFC from this paper.
Adds a parameter "balance_factor" to kmeans. Based on the paper's experimental results,
1.0seems to be a fairly decent default value, and we could consider making it the default in the future.This feature is restricted to the torch implementation for now. I think for it to work well, we'd actually have to incorporate it into the query path.
Later update: I'm actually not able to replicate the results from the paper at the moment, at least on the datasets I've tried.