lmcinnes / lmcinnes/umap

Support for wasserstein (earth mover's) distance

Open
#751 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
8.3k
Forks
871
Avg merge
1d 13h
Merged PRs (30d)
5

Description

On a longer term front, pynndescent, which is going to be the engine for nearest neighbor search in umap v0.5 (and was spun out of the original umap code), will have a wasserstein/kantorovich distance in version 0.5. A sparse/online version of that would probably be enough to implement word-mover distance directly in the nearest neighbor search itself. That is still a little way off however.

Originally posted by @lmcinnes in https://github.com/lmcinnes/umap/issues/424#issuecomment-625469698

What is the status on this? I'm using vector-based (not text-based) earth mover's distance.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the current nearest-neighbor distance support in UMAP and the referenced pynndescent work planned for version 0.5. Determine whether vector-based Wasserstein or Kantorovich distance is now available and how a sparse or online variant would fit the nearest-neighbor search. Done means the requested distance is supported for vector data with documented behavior and coverage in the relevant tests.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.