Scalable Second-order Riemannian Optimization for $K$-means Clustering

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Peng, Hou, Chun-Ying, Chen, Xiaohui, Zhang, Richard Y.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915832618024960
author Xu, Peng
Hou, Chun-Ying
Chen, Xiaohui
Zhang, Richard Y.
author_facet Xu, Peng
Hou, Chun-Ying
Chen, Xiaohui
Zhang, Richard Y.
contents Clustering is a hard discrete optimization problem. Nonconvex approaches such as low-rank semidefinite programming (SDP) have recently demonstrated promising statistical and local algorithmic guarantees for cluster recovery. Due to the combinatorial structure of the $K$-means clustering problem, current relaxation algorithms struggle to balance their constraint feasibility and objective optimality, presenting tremendous challenges in computing the second-order critical points with rigorous guarantees. In this paper, we provide a new formulation of the $K$-means problem as a smooth unconstrained optimization over a submanifold and characterize its Riemannian structures to allow it to be solved using a second-order cubic-regularized Riemannian Newton algorithm. By factorizing the $K$-means manifold into a product manifold, we show how each Newton subproblem can be solved in linear time. Our numerical experiments show that the proposed method converges significantly faster than the state-of-the-art first-order nonnegative low-rank factorization method, while achieving similarly optimal statistical accuracy.
format Preprint
id arxiv_https___arxiv_org_abs_2509_21675
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Scalable Second-order Riemannian Optimization for $K$-means Clustering
Xu, Peng
Hou, Chun-Ying
Chen, Xiaohui
Zhang, Richard Y.
Machine Learning
Optimization and Control
Clustering is a hard discrete optimization problem. Nonconvex approaches such as low-rank semidefinite programming (SDP) have recently demonstrated promising statistical and local algorithmic guarantees for cluster recovery. Due to the combinatorial structure of the $K$-means clustering problem, current relaxation algorithms struggle to balance their constraint feasibility and objective optimality, presenting tremendous challenges in computing the second-order critical points with rigorous guarantees. In this paper, we provide a new formulation of the $K$-means problem as a smooth unconstrained optimization over a submanifold and characterize its Riemannian structures to allow it to be solved using a second-order cubic-regularized Riemannian Newton algorithm. By factorizing the $K$-means manifold into a product manifold, we show how each Newton subproblem can be solved in linear time. Our numerical experiments show that the proposed method converges significantly faster than the state-of-the-art first-order nonnegative low-rank factorization method, while achieving similarly optimal statistical accuracy.
title Scalable Second-order Riemannian Optimization for $K$-means Clustering
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2509.21675