Hölder-Type Global Error Bounds for Non-degenerate Polynomial Systems

Abstract

Let F := (f 1, …, f p ): ℝ n → ℝ p be a polynomial map, and suppose that S := {x ∈ ℝ n : f i (x) ≤ 0,i = 1, …, p}≠∅. Let d := maxi =1, …, p deg f i and $\mathcal {H}(d, n, p) := d(6d - 3)^{n + p - 1}.$ Under the assumptions that the map F : ℝ n → ℝ p is convenient and non-degenerate at infinity, we show that there exists a constant c > 0 such that the following so-called Holder-type global error bound result holds $c d(x,S) \le [f(x)]_{+}^{\frac {2}{\mathcal {H}(2d, n, p)}} + [f(x)]_{+} \quad \textrm { for all } \quad x \in \mathbb {R}^{n},$ where d(x,S) denotes the Euclidean distance between x and S, f(x) := maxi=1, …, p f i (x), and [f(x)]+ := max{f(x),0}. The class of polynomial maps (with fixed Newton polyhedra), which are non-degenerate at infinity, is generic in the sense that it is an open and dense semi-algebraic set. Therefore, Holder-type global error bounds hold for a large class of polynomial maps, which can be recognized relatively easily from their combinatoric data. This follows up the result on a Frank-Wolfe type theorem for non-degenerate polynomial programs in Dinh et al. (Mathematical Programming Series A, 147(16), 519–538, 2014).