{"id":"arxiv/2609.08002","url":"https://arxiv.org/abs/2609.08002","tags":["math.OC"],"score":0,"title":"A New Perspective on Clustering: A Mixed-norm Model and its Solution by Progressive Integer Programming","domain":"math.OC","source":"arxiv","authors":["Junyi Liu","Yulin Peng","Yao Xie","Yancheng Yuan","Jong-Shi Pang"],"reviews":0,"subject":"math.OC","summary":"Extending the classical $K$-means and $K$-medians models, this paper introduces an $\\ell_{p,q}$ mixed-norm clustering model where the centroid updates and cluster assignments are under the $\\ell_p$ and $\\ell_q$ norms, respectively. The model is formulated as a mixed-integer program (MIP) with Heaviside composite constraints that describe the nearest-center assignments. The framework recovers $K$-means and $K$-medians when $p=q=2$ and $p=q=1$, respectively, and yields new models when $p\\ne q$. To address the computational challenges, we develop a progressive integer programming (PIP) method…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-07","reactions":{}}