Dimensionality Reduction Meets Network Science: Enhancing Data Sensemaking with UMAP's Internal kNN Graph

Forum topic · 小凯 · 2026-07-13

Summary

A 2025 arXiv paper (2507.08728) by Duen Horng Chau, Donghao Ren, and Fred Hohman proposes tapping into the k-nearest-neighbor (kNN) graph that UMAP constructs internally, which is typically discarded in standard workflows. Since this graph encodes the data manifold in the original high-dimensional space before UMAP's 2D projection introduces distortion, applying standard network science algorithms to it can enhance data sensemaking. The authors show that PageRank identifies representative data points, k-core decomposition reveals dense core regions versus sparse periphery, and clustering coefficients detect tight-knit neighborhoods of highly similar points. Quantitative and qualitative evaluations on MNIST and Fashion MNIST demonstrate these graph-based analyses are competitive with, or complementary to, purpose-built methods such as k-medoids for exemplar selection and HDBSCAN for density-based clustering.

Overview

Field: Machine Learning
Authors: Duen Horng Chau, Donghao Ren, Fred Hohman
Published: 2025-07-12
arXiv: 2507.08728

Summary

While UMAP is widely used for exploring high-dimensional data, typical workflows focus on its lower-dimensional embedding, largely overlooking the rich k-nearest-neighbor (kNN) graph that UMAP constructs internally. This graph encodes the data manifold in its original high-dimensional space, before the distortion that UMAP's 2D projection introduces.

The paper demonstrates the untapped potential of this internal representation, showing how standard graph algorithms applied to this graph enhance data sensemaking:

1. PageRank identifies representative data points. 2. k-core decomposition reveals dense core regions versus sparse periphery. 3. Clustering coefficient detects tight-knit neighborhoods with highly-similar data points.

Through quantitative and qualitative evaluation on MNIST and Fashion MNIST, the authors show that these graph-based analyses are not only practical but also competitive with or complementary to purpose-built methods such as k-medoids for exemplar selection and HDBSCAN for density-based clustering.

*Auto-collected on 2026-07-13*

Dimensionality Reduction Meets Network Science: Enhancing Data Sensemaking with UMAP's Internal kNN Graph

Summary

Overview

Summary

Tags