1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
# src/kmeans ## Purpose K-means++ clustering: smart initialization + iterative Lloyd refinement. Used by spectral clustering backend. ## Surfaces - `kmeans_pp(embeddings, k, max_iter) -> Vec<usize>` ## Dependencies None (self-contained). ## Invariants - Output labels are in 0..k. ## Verification ```bash cargo test --lib kmeans ``` ## Notes - Uses Euclidean distance on f32 vectors.