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Outdated suggestions cannot be applied.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.
ParlayANN implements a variety of algorithms as described in the paper: https://dl.acm.org/doi/10.1145/3627535.3638475 This pull request uses the ParlayANN version of Vamana (with some significant changes from the original). It was designed for large data sets and high parallelism, but works well on smaller graphs and a single thread.
It does not use any product quantization, but does use scalar quantization, and for large dimensions (just GIST is large enough for this suite) uses the Johnson-Lindenstrauss transform to reduce dimensionality.