Uniform Continuity in Distribution for Borel Transformations of Random Fields
Alexander I. Bufetov · 2025
Read Uniform Continuity in Distribution for Borel Transformations of Random Fields by Alexander I. Bufetov (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-24441.txt.
Introduction
Alexander Bufetov gives simple conditions for a continuity property of random fields. A random field assigns random values across a set of sites. Borel maps reshape those values. The question is when the reshaped field's law varies in a uniform, continuous way with the input. Bufetov's conditions ensure exactly that, in a brief note for readers in probability theory.
Catalog record
| CREATOR | Alexander I. Bufetov |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-24441
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Uniform Continuity in Distribution for Borel Transformations of Random Fields Alexander I. Bufetov Abstract Simple sufficient conditions are given that ensure the uniform continuity in distribution for Borel transformations of random fields. Моему дорогому научному руководителю Якову Григорьевичу Синаю с любовью и благодарностью. 1 Introduction and the statement of the main result Let W be a complete separable metric space, let {\mathfrak{M}}_{1}(W) be the space of Borel probability measures on W endowed with the weak topology and let {\mathfrak{N}} be a compact subset of the space {\mathfrak{M}}_{1}(W) . Assume that V\subset W is a Borel subset satisfying \eta(V)=1 for any measure \eta\in{\mathfrak{N}} . A Borel mapping g\colon{\mathfrak{N}}\times V\to W (1) induces a Borel mapping g_{*}\colon{\mathfrak{N}}\to{\mathfrak{M}}_{1}(W) (2) by the formula g_{*}\eta=(g(\eta,\cdot))_{*}\eta. (3) The map g_{*} is equivalently given by the formula \int_{W}\varphi\,dg_{*}\eta=\int_{W}\varphi(g(\eta,w))\,d\eta(w) (4) that holds for any bounded continuous function \varphi\colon W\to\mathbb{C} . The aim of this note is to give convenient sufficient conditions ensuring that the correspondence g\mapsto g_{*} induces a uniformly continuous map. Of particular interest to us is the case where the space W is the space of realizations of a random field; under additional assumptions on the regularity of the random field, the space W is a complete separable metric space (see Section 4 below). The space {\mathfrak{M}}_{1}(W) can be turned into a metric space in several different natural ways. For our purposes the Lévy–Prokhorov metric is particularly convenient, and we recall that the weak topology on {\mathfrak{M}}_{1}(W) is induced by the Lévy–Prokhorov metric d_{\mathrm{LP}} , which is uniquely defined by the following requirement (cf. Bogachev [ 5 ] , Billingsley [ 3 ] ): the Lévy–Prokhorov distance between measures \eta_{1},\eta_{2}\in{\mathfrak{M}}_{1}(W) does not exceed \varepsilon if and only if for every closed set K\subset W and its \varepsilon -neighbourhood K_{\varepsilon}=\{w\in W:d(w,K)\leq\varepsilon\} we have \eta_{1}(K)\leq\eta_{2}(K_{\varepsilon})+\varepsilon,\qquad\eta_{2}(K)\leq\eta_{1}(K_{\varepsilon})+\varepsilon. We endow the space of Borel mappings ( 1 ) with the metric of uniform convergence in probability, and the space of Borel mappings from {\mathfrak{N}} to \widetilde{{\mathfrak{N}}} with the Tchebycheff uniform metric. We fix a compact set \widetilde{{\mathfrak{N}}}\subset{\mathfrak{M}}_{1}(W) and only consider maps g of the form ( 1 ) satisfying the additional requirement g_{*}({\mathfrak{N}})\subset\widetilde{{\mathfrak{N}}} . The correspondence g\mapsto g_{*} is then uniformly continuous with respect to our metrics. We proceed to the precise formulations. We fix a metric d on the space W , and we write d_{\mathrm{LP}} for the corresponding Lévy–Prokhorov metric on the space {\mathfrak{M}}_{1}(W) . For a compact subset {\mathfrak{N}}\subset{\mathfrak{M}}_{1}(W) and a Borel subset V\subset W satisfying \eta(V)=1 for any \eta\in{\mathfrak{N}} , introduce the space \mathscr{B}({\mathfrak{N}}\times V,W) of Borel maps g\colon{\mathfrak{N}}\times V\to W. We endow the space \mathscr{B}({\mathfrak{N}}\times V,W) with the topology {\mathfrak{T}}_{\mathrm{prob}} generated by the neighbourhoods U(g_{0},\varepsilon,\delta)=\left\{g\in\mathscr{B}({\mathfrak{N}}\times V,W)\,:\,\eta(\{w\in W:d(g(\eta,w),g_{0}(\eta,w))>\varepsilon\})<\delta\text{ for all }\eta\in{\mathfrak{N}}\right\} for all \varepsilon>0 and \delta>0 . The collection of the sets U(g_{0},\varepsilon,\delta) over all \varepsilon>0,\delta>0 forms a subbasis for the topology {\mathfrak{T}}_{\mathrm{prob}} . The topology {\mathfrak{T}}_{\mathrm{prob}} on the space \mathscr{B}({\mathfrak{N}}\times V,W) is metrizable: one directly verifies that the distance d(g_{1},g_{2})=\sum_{l=1}^{\infty}2^{-l}\frac{\displaystyle\sup_{\eta\in{\mathfrak{N}}}\eta(\{w\in W:d(g_{1}(\eta,w),g_{2}(\eta,w))>2^{-l}\})}{1+\displaystyle\sup_{\eta\in{\mathfrak{N}}}\eta(\{w\in W:d(g_{1}(\eta,w),g_{2}(\eta,w))>2^{-l}\})} induces the topology {\mathfrak{T}}_{\mathrm{prob}} on \mathscr{B}({\mathfrak{N}}\times V,W) . Now let \mathscr{B}({\mathfrak{N}},{\mathfrak{M}}_{1}(W)) stand for the space of all Borel maps h\colon{\mathfrak{N}}\to{\mathfrak{M}}_{1}(W) . We endow the space \mathscr{B}({\mathfrak{N}},{\mathfrak{M}}_{1}(W)) with the Tchebycheff uniform metric with respect to the Lévy–Prokhorov metric on the space {\mathfrak{M}}_{1}(W) . For a compact set \widetilde{{\mathfrak{N}}}\subset{\mathfrak{M}}_{1}(W) let \mathscr{B}({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) be the subspace of Borel maps h\colon{\mathfrak{N}}\to{\mathfrak{M}}_{1}(W) satisfying the inclusion h({\mathfrak{N}})\subset\widetilde{{\mathfrak{N}}} . We note that the subspace \mathscr{B}({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) is closed in the Tchebycheff uniform topology. Let \widetilde{{\mathfrak{N}}}\subset{\mathfrak{M}}_{1}(W) be a compact set. Let \mathscr{B}_{\widetilde{{\mathfrak{N}}}}({\mathfrak{N}}\times V,W)\subset\mathscr{B}({\mathfrak{N}}\times V,W) be the subspace of maps g such that the map g_{*} defined by ( 3 ) satisfies the inclusion g_{*}({\mathfrak{N}})\subset\widetilde{{\mathfrak{N}}} . The subspace \mathscr{B}_{\widetilde{{\mathfrak{N}}}}({\mathfrak{N}}\times V,W)\subset\mathscr{B}({\mathfrak{N}}\times V,W) is closed by definition. The main result of this paper is Theorem 1.1 . The correspondence g\mapsto g_{*} induces a uniformly continuous map from the space \mathscr{B}_{\widetilde{{\mathfrak{N}}}}({\mathfrak{N}}\times V,W) to the space \mathscr{B}({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) . 2 Reduction to the case W=\mathbb{C}^{\mathbb{N}} Endow the space \mathbb{C}^{\mathbb{N}}=\{t=(t_{1},t_{2},\ldots):t_{n}\in\mathbb{C},\,n\in\mathbb{N}\} with the distance d(t^{(1)},t^{(2)})=\sum_{l=1}^{\infty}2^{-l}\frac{|t^{(1)}_{l}-t^{(2)}_{l}|}{1+|t^{(1)}_{l}-t^{(2)}_{l}|}. The Stone–Weierstrass Theorem directly implies the following two propositions. Proposition 2.1 . Let W be a complete separable metric space and let K be a compact subset of W . There exists a countable family of 1 -Lipschitz functions \Phi=(\varphi_{n})_{n\in\mathbb{N}} on K such that the correspondence w\mapsto(\varphi_{n}(w))_{n\in\mathbb{N}} defines a bi-Lipschitz homeomorphism of K onto its image. Proposition 2.2 . Let V and W be complete separable metric spaces. Let V^{\prime}\subset V and {\mathfrak{N}}\subset{\mathfrak{M}}_{1}(V) be Borel subsets such that \eta(V^{\prime})=1 for any \eta\in{\mathfrak{N}} . Let \tau\colon V^{\prime}\to W be a Borel map. Assume that for any \varepsilon>0 there exist R\geq 1 and a Borel subset V(\varepsilon)\subset V^{\prime} such that \sup_{\eta\in{\mathfrak{N}}}\eta(V^{\prime}\setminus V(\varepsilon))<\varepsilon, and for any x,y\in V(\varepsilon),x\neq y we have R^{-1}\leq\frac{d_{W}(\tau x,\tau y)}{d_{V}(x,y)}\leq R. Then the induced map \tau_{*}\colon{\mathfrak{N}}\to{\mathfrak{M}}_{1}(W) is a homeomorphism onto its image. Propositions 2.1 and 2.2 imply that it suffices to establish Theorem 1.1 for W=\mathbb{C}^{\mathbb{N}} . We now assume that {\mathfrak{N}} and \widetilde{{\mathfrak{N}}} are compact subsets of the space {\mathfrak{M}}_{1}(\mathbb{C}^{\mathbb{N}}) of Borel probability measures on the space \mathbb{C}^{\mathbb{N}}=\{t=(t_{1},t_{2},\ldots):t_{n}\in\mathbb{C},\,n\in\mathbb{N}\}. For a Borel subset V\subset\mathbb{C}^{\mathbb{N}} satisfying \eta(V)=1 for every \eta\in{\mathfrak{N}} , denote by \mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) the corresponding space of Borel maps g\colon{\mathfrak{N}}\times V\to\mathbb{C}^{\mathbb{N}},\qquad g(\eta,t)=(g^{(n)}(\eta,t))_{n\in\mathbb{N}}, where g^{(n)}(\eta,t)\in\mathbb{C} for n\in\mathbb{N} . The topology on \mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) is induced by the distance function d(g,\widehat{g})=\sum_{l,k=1}^{\infty}2^{-l-k}\frac{\displaystyle\sup_{\eta\in{\mathfrak{N}}}\eta(\{t\in V:|g^{(k)}(\eta,t)-\widehat{g}^{(k)}(\eta,t)|>2^{-l}\})}{1+\displaystyle\sup_{\eta\in{\mathfrak{N}}}\eta(\{t\in V:|g^{(k)}(\eta,t)-\widehat{g}^{(k)}(\eta,t)|>2^{-l}\})}. By definition, a sequence g^{(n)}\in\mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) converges to g if and only if for every k\in\mathbb{N} and every \varepsilon>0 we have \lim_{n\to\infty}\sup_{\eta\in{\mathfrak{N}}}\eta(\{t\in V:|g^{(n)}_{k}(\eta,t)-g^{\phantom{(n)}}_{k}(\eta,t)|>\varepsilon\})=0. In other words, the distance d induces the topology of coordinatewise convergence in probability, uniform in \eta\in{\mathfrak{N}} . As above, a Borel mapping g\in\mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) (cf. ( 4 )) induces a map g_{*}\colon{\mathfrak{N}}\to{\mathfrak{M}}_{1}(\mathbb{C}^{\mathbb{N}}) given by the formula g_{*}\eta=(g(\eta,\cdot))_{*}\eta. Recall that the space \mathscr{B}_{\widetilde{{\mathfrak{N}}}}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) is by definition the subspace of \mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) consisting of those maps g whose corresponding induced map g_{*} satisfies the inclusion g_{*}({\mathfrak{N}})\subset\widetilde{{\mathfrak{N}}} . As before, we endow the space \mathscr{B}({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) with the Tchebycheff uniform metric with respect to the Lévy–Prokhorov metric on {\mathfrak{M}}_{1}(\mathbb{C}^{\mathbb{N}}) . The subspace \mathscr{B}_{\widetilde{{\mathfrak{N}}}}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) is closed in \mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) . The special case of Theorem 1.1 in our particular setting is the following Theorem 2.3 . The correspondence g\mapsto g_{*} defines a uniformly continuous map from \mathscr{B}_{\widetilde{{\mathfrak{N}}}}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) to \mathscr{B}({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) . As we have seen, Theorem 2.3 directly implies Theorem 1.1 . Corollary 2.4 . Let Borel maps g^{(n)}\in\mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) , n\in\mathbb{N} , be such that the map (g^{(n)})_{*} is continuous for every n\in\mathbb{N} and g^{(n)}\to g in \mathscr{B}({\mathfrak{N}}\times V,\mathbb{C}^{\mathbb{N}}) , then g_{*} is uniformly continuous. Let a subset S_{0}\subset\mathbb{N} and a compact subset {\mathfrak{N}}\subset{\mathfrak{M}}_{1}(\mathbb{C}^{\mathbb{N}}) satisfy the following condition: for every n\in\mathbb{N} and every \delta>0,\varepsilon>0 there exists s_{0}\in S_{0} such that \sup_{\eta\in{\mathfrak{N}}}\eta(\{t\in\mathbb{C}^{\mathbb{N}}:|t(n)-t(s_{0})|>\delta\})<\varepsilon. (5) The above argument implies the following Proposition 2.5 . Under condition ( 5 ), the projection \mathbb{C}^{\mathbb{N}}\to\mathbb{C}^{S_{0}} induces a homeomorphism of the compact set {\mathfrak{N}} onto its image. 3 Proof of Theorem 2.3 Let {\mathfrak{N}} and \widetilde{{\mathfrak{N}}} be compact subsets of the space {\mathfrak{M}}_{1}(\mathbb{C}^{\mathbb{N}}) . Endow the space C({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) of continuous maps from {\mathfrak{N}} to \widetilde{{\mathfrak{N}}} with the Tchebycheff metric: for h,h^{\prime}\in C({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) the Tchebycheff uniform distance d(h,h^{\prime}) is defined by the formula d(h,h^{\prime})=\max_{\eta\in{\mathfrak{N}}}d_{\mathrm{LP}}\left(h(\eta),h^{\prime}(\eta)\right). The space C({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) is a complete separable metric space. To a map h\in C({\mathfrak{N}},\widetilde{{\mathfrak{N}}}) we assign its projection onto the first l coordinates \pi_{l}h(\kappa)=\operatorname{distr}^{l}_{h(\kappa)}([1,l]), where the symbol \operatorname{distr}^{l}_{\eta} stands for the l -dimensional marginal distribution of a measure \eta on the first l coordinates. Proposition 3.1