{"id":"arxiv/2601.13915","url":"https://arxiv.org/abs/2601.13915","tags":["math.CA","math.FA"],"score":0,"title":"Geometry-Driven Conditioning of Multivariate Vandermonde Matrices in High-Degree Regimes","domain":"math.CA","source":"arxiv","authors":["Omer Friedland","Yosef Yomdin"],"reviews":0,"subject":"math.CA","summary":"We study multivariate monomial Vandermonde matrices $V_N(Z)$ with arbitrary distinct nodes $Z=\\{z_1,\\dots,z_s\\}\\subset B_2^n$ in the high-degree regime $N\\ge s-1$. Introducing a projection-based geometric statistic -- the \\emph{max-min projection separation} $ρ(Z,j)$ and its minimum $κ(Z)=\\min_jρ(Z,j)$ -- we construct Lagrange polynomials $Q_j\\in\\mathcal P_N^n$ with explicit coefficient bounds $$ \\|Q_j\\|_\\infty \\lesssim s\\Bigl(\\frac{4n}{ρ(Z,j)}\\Bigr)^{s-1}. $$ These polynomials yield quantitative distance-to-span estimates for the rows of $V_N(Z)$ and, as consequences, $$ σ_{\\min}(V_N(Z))…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-01-20","reactions":{}}