Why does clustering degrade in high dimensions?
mediumAnswer
- In high dimensions, pairwise distances concentrate — points look almost equidistant — so distance-based clustering loses discriminative power.
- Also, irrelevant features add noise.
- Fixes: dimensionality reduction (PCA/UMAP) before clustering, feature selection, or use models less reliant on Euclidean distance (e.g., subspace clustering).
Check yourself — multiple choice
- Distances become more informative as dimensionality grows
- Distances concentrate and clustering becomes less discriminative
- PCA cannot help
- Adding noise features improves clustering
Distance concentration in high dimensions weakens neighborhood-based methods.
#clustering#dimensionality-reduction
Practise Unsupervised Learning
214 interview questions in this topic.