lmcinnes / lmcinnes/umap

Can GNNs be used as encoders in parametric UMAP ?

Open
#905 0 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

Having gone through the documentation of parametric UMAP, I thought we could use graph neural networks to get the embedding of the KNN graph in a lower dimension. Is it possible to access the KNN graph(the one created in the first step of UMAP) in a higher dimension and then use GNN to update each node's embedding so as to capture neighborhood information and then reduce that updated embedding to a lower dimension using a fully connected layer?

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 parametric UMAP documentation and the first-step KNN graph described in the issue. Determine whether that graph is accessible for a GNN encoder and define a concrete implementation scope and acceptance criteria before proposing work.

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.