Skip to main content

> ML_ALGORITHM // UNIFORM-MANIFOLD-APPROXIMATION-PROJECTION_v1.0

Uniform Manifold Approximation and Projection (UMAP)

Fast manifold learning technique based on topological data analysis that preserves both local and global data structure with parametric projection.

Non-linear Manifold Learningclassical-unsupervisedmoderate-posthocmedium (1k-100k)
Back to All Algorithms
Computational Complexity
Training Complexity:O(n * log n) via nearest neighbor descent
Inference Complexity:O(log n) out-of-sample projection supported
Hardware Profile
CPU Friendly:Yes
Requires GPU:No
Memory Footprint:moderate
Interpretability & Data
Interpretability Tier:moderate-posthoc
Training Data Needs:medium (1k-100k)

Interpretability Assessment

Preserves significantly more global topological structure than t-SNE while maintaining tight local clusters.

Suitable Tasks & Supported Modalities

Suitable Tasks:
dimensionality reductiondata visualizationfeature extraction
Supported Modalities:
tabularembeddingsimage

Implementing Libraries

umap-learn
scikit-learnscikit-learn Consortium / Inria · v1.5.2
View Spec

Foundational Literature

UMAP: Uniform Manifold Approximation and Projection for Dimension ReductionLeland McInnes, John Healy (2018) · arXiv preprint
Common Pitfalls & Warnings
  • Assuming Euclidean geometry across embedded clusters without verifying metric compatibility