jlmelville/uwot

What it solves

It provides an R implementation of the Uniform Manifold Approximation and Projection (UMAP) method, which is used for dimensionality reduction of high-dimensional data into a lower-dimensional space for visualization or analysis.

How it works

uwot is a translation of the original Python UMAP implementation. It utilizes C++ code for performance and can integrate with optional packages like RcppHNSW and rnndescent to accelerate nearest neighbor searches, which are critical to the UMAP process. It also supports supervised learning and metric learning for out-of-sample extensions.

Who it’s for

Data scientists and researchers using the R programming language who need to perform dimensionality reduction on complex datasets.

Highlights

• UMAP Implementation: A faithful R port of the original Python UMAP method.
• C++ Backend: Uses compiled C++ for efficiency.
• Flexible Search: Supports multiple nearest neighbor methods, including HNSW and nndescent.
• Extended Capabilities: Includes supervised and metric learning extensions for out-of-sample data.