lmcinnes / lmcinnes/umap

UMAP and Locality Weighting

Open
#267 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

In the docs you say

> And our results in terms of adjusted Rand score and adjusted mutual information are in line with the current state of the art techniques using convolutional autoencoder techniques. That’s not bad for an approach that is simply viewing the data as arbitrary 784 dimensional vectors.

In your opinion, would it be possible to add a concept of "locality" to UMAP? So for an image, you could give more weight to features (pixels) that are "spatially" close to each other

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 with the documentation passage quoted in the issue and review how UMAP currently handles image data represented as 784-dimensional vectors. Define whether spatially local pixel weighting should be supported and what evaluation would demonstrate that it works; the issue does not name implementation files or 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.