Random linear functionals arising in stochastic integration

The author defines hypermeasures on Polish spaces and proves some properties of these measures. Notation: \(( \Omega, \mathcal{F}, P)\) is a complete probability space, \( (X, \rho)\) a Polish space, \(BL=\) the class of all bounded Lipchitz functions on \(X\), \(\mathcal{X}\) all Borel subsets of \(X\), and \(M\) all charges on \(\mathcal{X}\). For \(q \in M\), set \(\| q \|_{\rho}= \sup\{q(\varphi): \varphi \in BL\), \(\|\varphi \| \leq 1\), \(|\varphi (x)- \varphi (x)|\leq \rho (x,y) \}\). If \(BL\) separates the points of \(X\), then \(\| \cdot \|_{\rho}\) is a norm on \(M\), and if, in addition, \(X\) is infinite, then \(M\) is incomplete with this norm. The elements of the completion of \(M\), defined on \(BL\), are called hypermeasures. For \(n\in\mathbb N\), let \(\underline{n}=\{1,2,\dots,n\}\). For a sequence \(\{\chi_{n}\}\) of real-valued functions on a topological space \(Y\), \(n\in\mathbb N\), and \(H\subset \underline{n}\), \(g_n^H= \prod_{i\in H}\chi_i\prod_{i\in\underline{n}\setminus H} (1-\chi_{i})\); \(D_{n}^{H}\) will denote the \(\operatorname{supp}g_{n}^{H}\). This sequence is called a tile if: (i) for all \(n\), \(0 \leq \chi_{n}\leq 1\), (ii) for any tuple \((i_{1},\dots,i_{n})\in\{0,1\}^{n}\), the set \(Q_{i_{1},\dots,i_{n}}=\{y:\chi_{j}(y)=i_{j},\;j \in \underline{n} \}\) has nonempty interior, (iii) all \(\chi_{n}\), except at most finitely many, have sequentially compact support. A random linear functional \(\gamma\) on \(BL\) is called stochastically continuous if \(\gamma\varphi_{n}\overset{P}{\to}0\) for any uniformly bounded, pointwise convergent to \(0\) sequence \(\{\varphi_{n}\}\) of Lipschitz functions. A tile is called Lipschitz if all its members are Lipschitz functions. The main results of the present paper are as follows. I. Suppose that \((X,\rho)\) is a Polish space and \(\text{BL}\) separates the points of \(X\). Then, for any hypermeasure \(\nu\) and any uniformly bounded, pointwise convergent to \(0\) sequence \(\{\varphi_{n}\}\subset\text{BL}\) satisfying the condition that \(\sup_{n}\sup_{x\neq y}\frac{|\varphi_{n}(x)-\varphi_{n}(y)|}{\rho(x,y)}<\infty\), we have that \(\nu(\varphi_{n})\to 0\). II. Let \(\{\chi_{n}\}\) be a tile on a topological space \(Y\). Then there is a probability Baire measure \(m\) on \(Y\) with the property that for any \(n\in\mathbb N\) and \(H\subset\underline{n}\), \(\int g_n^H\,dm=2^{-n}\). III. Suppose that \((X,\rho)\) is a Polish space, with \(\rho\) bounded and \(\gamma\) a stochastically continuous random linear functional on \(\text{BL}(X,\rho)\). Assume further that \(\Phi=\{\varphi\in\text{BL}:\|\varphi\| \leq 1,\;|\varphi(x)-\varphi(x)|\leq\rho(x,y)\}\) is separable in Tikhonov's topology and there exists a Lipschitz tile \(\{\chi_{n}\}\) such that \(\sup_{n\in\mathbb N,\;H\subset \underline{n}}(2^{n}\operatorname{diam}D_{n}^{H})<\infty\) and \(\sup_{n}E(\sum_{H \subset\underline{n}}(\gamma g_{n}^{H})^{2})^{\frac{1}{2}}<\infty\). Then there exists a random hypermeasure \(\nu\) such that for any \(\varphi\in\text{BL}\), \(\gamma \varphi = \nu \varphi\) a.s. Some corollaries and related additional results are also proved.

Recommendations

Cites work

Cited in

(9)

This page was built for publication: Random linear functionals arising in stochastic integration

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q996753)