{"id":"arxiv/2611.00514","url":"https://arxiv.org/abs/2611.00514","tags":["math.ST"],"score":0,"title":"Unconstrained support points can leave a convex support","domain":"math.ST","source":"arxiv","authors":["Zhaohui Li"],"reviews":0,"subject":"math.ST","summary":"Support points provide a practical and effective way to approximate a probability distribution by a small set of equally weighted representatives through energy-distance minimization. In many applications, the variables satisfy convex constraints that the representatives should also respect. We investigate whether unconstrained support points automatically satisfy such constraints and show that this is not guaranteed. We construct counterexamples using three planar observations, a density with full triangular support, and positive-definite matrix data admitting an indefinite optimal…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-31","reactions":{}}