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Outdated suggestions cannot be applied.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 replaces
sparse_mutual_reachability
with a version which is 137 times faster for my datasets. The algorithm is essentially identical but it operates on the CSR matrix directly, omitting conversion to LIL and back which is both very slow and uses a lot of unnecessary memory.The code supports both float32 and float64 natively which allows for more memory savings if double precision is not required. Additionally we save another unnecessary copy by passing
overwrite=True
tocsgraph.minimum_spanning_tree
.Lastly the user can use
overwrite=True
when calling hdbscan to indicate the distance matrix they pass can be modified in place, saving yet another copy.Overall memory usage can be cut to 1/4th which makes it possible to deal with very large distance matrices. I have successfully clustered graphs with 10M nodes and 1B edges.