Prophet inequalities for finite stage multiparameter optimal stopping problems (Q2506420)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Prophet inequalities for finite stage multiparameter optimal stopping problems |
scientific article |
Statements
Prophet inequalities for finite stage multiparameter optimal stopping problems (English)
0 references
28 September 2006
0 references
The author deals with the optimal stopping problem for a stochastic field with the discrete multivariate time domain \(Z={\mathbb N}^d\cup\{\infty\}\). Herein, multiple play means that before stopping, at each stage the decision-maker has the option to proceed within \(Z\) from \(\sigma_t\) to \(\sigma_{t+1}:=\sigma_t+\vec{e}\) with \(\vec{e}\in\{0,1\}^d\setminus\{(0,0,\ldots,0)\}\). Let \((\Omega, {\mathcal F}, P)\) be a complete probability space equipped with a family \(\{{\mathcal F}_z:z\in {\mathbb N}^d\}\) of sub-\(\sigma\)-fields of \({\mathcal F}\) which satisfies the following conditions: \({\mathcal F}_0\) contains all \(P\)-null sets of \({\mathcal F}\), and if \(z\leq w\), then \({\mathcal F}_z\subset {\mathcal F}_w\). An \(\{{\mathcal F}_z\}\)-stopping point is a random variable \(T\) taking values in \(J\) such that \(\{T\leq z\}\in {\mathcal F}_z\) for all \(z\in J\), where \(J\) is the index set. When stopping, a terminal reward will be earned. A tactic is a pair \((\{\sigma_n\},\tau)\) with a sequence \(\{\sigma_n\}\) from \(Z\) having the following properties: \(\sigma(0)=0\), \(\sigma(n)\) is an \({\mathcal F}_z\)-stopping point for all \(n\), \(\sigma(n+1)\in D_{\sigma(n)}\) \(P\)-a.e., where \(D_z\) is the set of all direct successors of \(z\) and \(\sigma(n+1)\) is \({\mathcal F}_{\sigma(n)}\)-measurable for all \(n\). \(\tau\) is a common \(\{{\mathcal F}_{\sigma(n)}, n\in {\mathbb N}\}\)-stopping time. A stopping point \(T\) is called accessible if there exists a tactic \((\{\sigma(n)\},\tau)\) such that \(T=\sigma(\tau)\). In the classical optimal stopping theory [cf. \textit{Y. S.~Chow, H.~Robbins} and \textit{D.~Siegmund}, ``Great expectations: The theory of optimal stopping'' (1971; Zbl 0233.60044)], the comparisons between the expected reward of a player with complete foresight and the expected reward of the player using nonanticipating stop rules, known as prophet inequalities, have been studied by many authors. The case of independent \([0,1]\)-random variables has been considered by \textit{U.~Krengel} and \textit{L.~Sucheston} [On semiamarts, amarts, and processes with finite value. Probability on Banach spaces, in: Kuelbs, James (ed.), Advances in Probability and Related Topics. Vol. 4. New York, Basel: Marcel Dekker, Inc., 197--266 (1978)]. The review of the subject for the one-parameter optimal stopping problems can be found in the paper by \textit{T. P.~Hill} and \textit{R. P.~Kertz} [in: Strategies for sequential search and selection in real times. Contemp. Math. 125, 191--207 (1992; Zbl 0794.60040)]. The prophet inequalities in the case of multiparameter optimal stopping problems have been studied by \textit{U.~Krengel} and \textit{L.~Sucheston} [J. Multivariate Anal. 11, 199--229 (1981; Zbl 0461.60059)]. The paper is devoted to the prophet inequalities for optimal stopping of independent nonnegative multiparameter stochastic process indexed by \(I_n=\{z\in {\mathbb N}^d:| z| \leq n\}\) for \(n\in {\mathbb N}\) and \(I_u=\{z\in {\mathbb N}^d: | z| \leq u\}\) for \(u\in {\mathbb N}^d\). It is shown that \[ {\mathbf E}\Bigl[\max_{z\in I_n}\{X(z)\}\Bigr]<(p(n)+1)\sup_{T\in A(I_n)}{\mathbf E}[X(T)], \] where \(p(k)=\text{card}\{z:| z| =k, z\in I_n\}\), \(k\leq n\), and \(A(I_n)\) is the set of all accessible point in \(I_n\). It extends the inequality for finite stage two-parameter stopping problem from a paper by the author [``Prophet inequalities for two-parameter optimal stopping problems'']. It is also proved that \({\mathbf E}[\max_{z\in I_t}\{X(z)\}]<(q(m^\star)+1)\sup_{T\in A(I_t)}{\mathbf E}[X(T)]\), where \(q(k)=\text{card}\{z:| z| =k, z\in I_t\}\), \(k\leq | t| \) and \(m^\star=\min\{k:\max_{i\leq | t| }q(i)=q(k)\}\).
0 references
multi-parameter stochastic process
0 references
tactic
0 references
inequalities for stochastic processes
0 references