# UMAP

> Source: https://sukruyusufkaya.com/en/glossary/umap
> Updated: 2026-05-13T20:57:32.067Z
> Type: glossary
> Category: makine-ogrenmesi
**TLDR:** A modern and efficient dimensionality reduction technique that attempts to preserve both local and global data structure.

<p>UMAP has become a prominent method in recent years for high-dimensional visualization and feature compression. Compared with t-SNE, it can often run faster and may preserve more of the global structure. It has gained widespread use in biology, NLP, computer vision, and embedding analysis. Still, its outputs must be interpreted carefully, since a low-dimensional projection does not always represent the full structural truth of the data.</p>