BS\(\Delta\)Es and BSDEs with non-Lipschitz drivers: comparison, convergence and robustness (Q358147)

From MaRDI portal
Revision as of 12:50, 18 April 2024 by Importer (talk | contribs) (‎Changed an Item)
scientific article
Language Label Description Also known as
English
BS\(\Delta\)Es and BSDEs with non-Lipschitz drivers: comparison, convergence and robustness
scientific article

    Statements

    BS\(\Delta\)Es and BSDEs with non-Lipschitz drivers: comparison, convergence and robustness (English)
    0 references
    0 references
    0 references
    0 references
    16 August 2013
    0 references
    The authors study in their work the approximation of backward stochastic differential equations (BSDEs) with a driver \(f\) and a driving \(d\)-dimensional Brownian motion \(W\) by backward stochastic difference equations (BS\(\Delta\)Es) with driving coefficient \(f^N\) and a governing \(d\)-dimensional random walk; \(W^N\) approximates \(W\) and \(f^N\) converges to \(f\). The novelty of the authors' approach consists in the fact that the drivers \(f\) and \(f^N\) may be non-Lipschitz. They discuss conditions under which, for every bounded terminal condition \(\xi\) for the BSDE, there exist a bounded terminal value \(\xi^N\) and an associated solution for the BS\(\Delta\)E with driver \(f^N\), such that both \(\xi^N\) and the solution of this BS\(\Delta\)E converge to \(\xi\) and the associated solution of the BSDE. In the particular case where \(f\) and \(f^N\) are convex in \(z\), this convergence of the solutions turns out to hold true whenever \(\xi^N,\;N\geq1,\) is a uniformly bounded sequences converging to \(\xi\). The obtained results are based on a careful study of BS\(\Delta\)Es, which contains in particular a comparison principle for BS\(\Delta\)Es. With an example the authors show that, unlike to the existence of solutions of BSDEs whose driver \(f\) has a quadratic growth in \(z\), if one wants the convergence of approximation schemes with BS\(\Delta\)Es under rather general assumptions, the growth of the drivers has to be supposed strictly less than quadratic in \(z\), but it may be more than linear in \(z\). With their studies, the authors generalise former results of several authors considering Lipschitz drivers. Recent works by \textit{P. Imkeller} and \textit{G. dos Reis} [Stochastic Processes Appl. 120, No.~3, 348--379 (2010; Zbl 1196.60101); corrigendum ibid. 120, No.~11, 2286--2288 (2010; Zbl 1229.60071)] and by \textit{A. Richou} [Ann. Appl. Probab. 21, No. 5, 1933--1964 (2011; Zbl 1274.60221)] study the approximation for drivers with quadratic growth in \(z\), but under regularity assumptions on \(\xi\), \(\xi^N\) and specially chosen \(f^N\).
    0 references
    backward stochastic differential equation
    0 references
    backward stochastic difference equation
    0 references
    comparison theorem
    0 references
    approximation scheme
    0 references
    robustness
    0 references

    Identifiers