lmcinnes/umap

HNSW support

Open

#213 opened on Mar 25, 2019

View on GitHub
 (3 comments) (0 reactions) (0 assignees)Python (751 forks)batch import
help wanted

Repository metrics

Stars
 (6,478 stars)
PR merge metrics
 (Avg merge 2d 1h) (1 merged PR in 30d)

Description

Just curious if the HNSW algorithm is amenable to use in this project? Probably not directly because I see you are jitting in the distance metric but curious if it is worth a look into that algorithm for the nn search vs random projections? It is consistently the top performing algorithm on the ann shootout and I was wondering if the performance of UMAP is significantly impacted by the NN component and if so would incorporating HNSW make a meaningful difference?

Contributor guide

Research direction
Investigate whether the HNSW algorithm can replace the current nearest neighbor search in UMAP, considering the performance impact and compatibility with the jitted distance metric. Start by reviewing HNSW's data structure and comparing it to UMAP's existing approximate nearest neighbor search.
Tech stack
python
Domain
backendmachine learning
Issue type
Research
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Fresh
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Mostly clear
Prerequisites
Pythonmachine learning
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
30