umap-learn
Applies UMAP-learn to nonlinear dimensionality reduction, 2D/3D embeddings
by k-dense-aiFree
Description
umap-learn is an agent skill for UMAP, which stands for Uniform Manifold Approximation and Projection, a technique for non-linear dimensionality reduction. It is used for visualisation, for fast and scalable embeddings that preserve neighbourhood structure, for supervised and semi-supervised embeddings, and for preprocessing before clustering. It also covers variants such as DensMAP and AlignedUMAP. The skill tells you what to check before interpreting an embedding: that the coordinates are finite, that neighbourhoods are retained, that results are not overly sensitive to seeds and parameters, and that the domain evidence in the original space agrees. It warns that UMAP axes, island areas and gaps between clusters carry no calibrated physical or probabilistic meaning. It also says to keep predictive test data out of both preprocessing and the embedding fit. The current stable release, umap-learn 0.5.12, was verified on 1 October 2026. It needs Python 3.9 or newer, and depends on scikit-learn 1.6 or newer, numba, pynndescent, NumPy and SciPy. Pin the version for reproducible work.…
Highlights
- Non-linear dimensionality reduction
- 2D and 3D embeddings
- Supervised and semi-supervised UMAP
- DensMAP and AlignedUMAP
- Checks before interpreting an embedding
Ratings & Reviews
No ratings yet
Be the first to rate umap-learn.
People also visit
umap-learn FAQ
What is UMAP used for?
Visualisation and non-linear dimensionality reduction, with embeddings that preserve neighbourhood structure.
Do distances in the plot mean something?
The skill says UMAP axes, island areas and cluster gaps have no calibrated physical or probabilistic meaning.
What should I check before trusting an embedding?
Finite coordinates, neighbourhood retention, sensitivity to seeds and parameters, and agreement with evidence in the original space.
Which version is current?
umap-learn 0.5.12, the stable release verified on 1 October 2026.
How do I install it?
Run npx skills add https://github.com/k-dense-ai/scientific-agent-skills --skill umap-learn.
Is it checked for security?
Its security audits show passes from Gen Agent Trust Hub, Socket and Snyk.
What are the best umap-learn alternatives?
Popular alternatives to umap-learn include etetoolkit, songsee and data-scraper-agent. See all umap-learn alternatives