TheoremBase

Gaussian Smoothing of a Bounded Borel Function on Euclidean Space: the Two Forms, Derivatives through the Kernel and through the Function, and Pointwise Convergence

The convolution of a bounded Borel function with a Gaussian kernel is bounded by the same constant, continuously differentiable with derivatives computed through the kernel, commutes with differentiation when the function is bounded C1C^1, and converges to the function at its points of continuity.

Statement

In the setting of Probability Measures on Euclidean Space and Random Vectors: Standing Notation, with Differential Calculus and Convexity on Euclidean Open Sets: Standing Notation in force, fix a dimension qq, and let gsg_{s} for 0<s≤10<s\le1, and the notation s−1/2s^{-1/2}, be those of The Gaussian Kernels on Euclidean Space: Scaling, Derivatives up to Order Three, the Convolution Identity, Moments and Tails; each gsg_{s} is smooth on Rq\mathbb{R}^{q}, so that its partial derivatives ∂igs\partial_{i}g_{s} exist, by The Gaussian Kernels on Euclidean Space: Scaling, Derivatives up to Order Three, the Convolution Identity, Moments and Tails §derivatives. Lebesgue measure λq\lambda_{q} on B(Rq)\mathcal{B}(\mathbb{R}^{q}) is that of Euclidean Space and Lebesgue Measure: Standing Notation §measure, integrals with respect to it are also written ∫f(z) dz\int f(z)\,dz, and integrable means integrable with respect to λq\lambda_{q}. The class C1C^{1} on Rq\mathbb{R}^{q} and partial derivatives are those of Differential Calculus and Convexity on Euclidean Open Sets: Standing Notation, Rq\mathbb{R}^{q} being open by claim 1 of Euclidean Space is Open in Itself, and CkC^k Maps are Continuous, and Cb1(Rq)C^{1}_{b}(\mathbb{R}^{q}) is the set of Bounded Continuously Differentiable Functions with Bounded Partial Derivatives on Euclidean Space §bounded. Let H:Rq→RH:\mathbb{R}^{q}\to\mathbb{R} be Borel and let MM be a nonnegative real number with ∣H(z)∣≤M|H(z)|\le M for every z∈Rqz\in\mathbb{R}^{q}.

1. (Smoothing) Let 0<s≤10<s\le1 and y∈Rqy\in\mathbb{R}^{q}. The functions x↦H(y−x) gs(x)x\mapsto H(y-x)\,g_{s}(x) and z↦gs(y−z) H(z)z\mapsto g_{s}(y-z)\,H(z) on Rq\mathbb{R}^{q} are Borel and integrable, and their integrals are equal; their common value is written

Hs(y)=∫RqH(y−x) gs(x) dx=∫Rqgs(y−z) H(z) dz,H_{s}(y)=\int_{\mathbb{R}^{q}}H(y-x)\,g_{s}(x)\,dx=\int_{\mathbb{R}^{q}}g_{s}(y-z)\,H(z)\,dz ,

and ∣Hs(y)∣≤M|H_{s}(y)|\le M.

2. (Derivatives through the kernel) Let 0<s≤10<s\le1. The function Hs:Rq→RH_{s}:\mathbb{R}^{q}\to\mathbb{R} is of class C1C^{1} on Rq\mathbb{R}^{q}, and for every i∈[q]i\in[q] and every y∈Rqy\in\mathbb{R}^{q} the function z↦∂igs(y−z) H(z)z\mapsto\partial_{i}g_{s}(y-z)\,H(z) is Borel and integrable, with

∂i(Hs)(y)=∫Rq∂igs(y−z) H(z) dz,∣∂i(Hs)(y)∣≤M q  s−1/2.\partial_{i}(H_{s})(y)=\int_{\mathbb{R}^{q}}\partial_{i}g_{s}(y-z)\,H(z)\,dz,\qquad|\partial_{i}(H_{s})(y)|\le M\,\sqrt{q}\;s^{-1/2}.

3. (Derivatives through the function) Suppose that H∈Cb1(Rq)H\in C^{1}_{b}(\mathbb{R}^{q}), and let 0<s≤10<s\le1 and i∈[q]i\in[q]. Then ∂iH\partial_{i}H is Borel and bounded, so that its smoothing (∂iH)s(\partial_{i}H)_{s} is defined by claim 1 applied with ∂iH\partial_{i}H and a bound of it in place of HH and MM, and ∂i(Hs)=(∂iH)s\partial_{i}(H_{s})=(\partial_{i}H)_{s}.

4. (Pointwise convergence) Let y∈Rqy\in\mathbb{R}^{q} be a point at which HH is continuous. For every positive ε∈R\varepsilon\in\mathbb{R} there is s0∈Rs_{0}\in\mathbb{R} with 0<s0≤10<s_{0}\le1 such that ∣Hs(y)−H(y)∣<ε|H_{s}(y)-H(y)|<\varepsilon for every ss with 0<s≤s00<s\le s_{0}.

Proofs

Log in to submit a proof.

Loading...

Citations

Loading…

Dependencies

Loading…

Related

0 relations

Curated associations between results. These are editable and subjective — they do not replace the dependency graph, which is derived from the references in the text.

No relations recorded yet.

Comments

Log in to comment.

Loading…