hubness-skews-knn-indegree-in-high-dimensions
IN premise — summaries/2026/08/24/wiki-Curse_of_dimensionality.md
Created 2026-08-24T17:11:09+00:00
In k-NN digraphs in high dimensions, the indegree distribution becomes skewed such that a small set of 'hub' points appear disproportionately in neighbor lists, distorting classification and clustering.
Summary
In high-dimensional data, a handful of points end up showing up in the neighbor lists of almost everyone else, making the structure of the data look artificially centered on those few points. This means standard nearest-neighbor classification and clustering will be systematically biased, since those few "attractor" points distort what the true neighborhood relationships should be.