UMAP-Learn skill: what it does and how to install it

What it does. Applies the Python umap-learn estimator to scaled or metric-matched data, producing embeddings for visualisation, clustering preprocessing or downstream models. It covers supervised and semi-supervised fits, new-data transforms, HDBSCAN workflows, densMAP, Parametric UMAP, inverse transforms and AlignedUMAP, with guidance on tuning and stability checks.

When to use it. Use for nonlinear dimensionality reduction or neighbourhood-focused analysis of numeric, sparse, text, binary or related datasets. Do not treat axes, gaps or island sizes as calibrated quantities, and do not use densMAP for unseen-data transforms; validate embeddings against the original space and held-out data.