UMAP Overview

Linear Algebra for Machine Learning · Axiom Academy

Fast, scalable manifold learning for visualization and preprocessing UMAP: Modern Alternative to t-SNE UMAP stands for "Uniform Manifold Approximation and Projection." It's a newer algorithm that preserves both local and global structure, scales to millions of points, and is fast. Controls the size of local neighborhood (default: 15). Lower values = more local focus. Higher values = more global structure. Minimum distance between points in output space (default: 0.1). Lower values = tighter clusters. Higher values = more spread. Distance metric: euclidean, manhattan, cosine, etc. Choose based on your data type. Most libraries (Python: umap-learn, R: umap) make it easy to use:

This is the written version of the interactive lesson above. See the full Linear Algebra for Machine Learning course.