BSE's, BSDE's and fixed point problems

Patrick Cheridito, Kihun Nam

Abstract

In this paper, we introduce a class of backward stochastic equations (BSEs) that extend classical BSDEs and include many interesting examples of generalized BSDEs as well as semimartingale backward equations. We show that a BSE can be translated into a fixed point problem in a space of random vectors. This makes it possible to employ general fixed point arguments to find solutions. For instance, Banach's contraction mapping theorem can be used to derive general existence and uniqueness results for equations with Lipschitz coefficients, whereas Schauder-type fixed point arguments can be applied to non-Lipschitz equations. The approach works equally well for multidimensional as for one-dimensional equations and leads to results in several interesting cases such as equations with path-dependent coefficients, anticipating equations, McKean-Vlasov type equations and equations with coefficients of superlinear growth.

Introduction

Stochastic calculus often feels like a collection of specialized tools designed for specific, isolated problems. In their research, Patrick Cheridito and Kihun Nam take a significant step toward a more unified theory. By introducing a broad class of equations termed Backward Stochastic Equations (BSEs), they demonstrate that many complex problems in stochastic analysis can be solved by translating them into the universal language of fixed point theory.

Classical Backward Stochastic Differential Equations (BSDEs) are a cornerstone of modern mathematical finance and stochastic control. However, they often come with rigid constraints: the underlying noise is typically restricted to Brownian motion, and the "driver" function is often required to depend only on the current state of the process. Real-world applications demand more flexibility, including path-dependency, anticipating drivers, and McKean–Vlasov dynamics. The BSE framework provides a unified method to handle these diverse scenarios under one theoretical roof.

The BSE Framework: From Processes to Fixed Points

The central innovation of this work is the realization that a BSE can be viewed as a mapping between random vectors rather than just a differential relationship over time.

In a standard BSDE, one seeks a pair of processes $(Y, Z)$ satisfying a stochastic integral equation. Cheridito and Nam simplify this by focusing on the terminal value or a related random vector. They translate the BSE into a problem of finding a single random vector $V$ such that: $$V = G(V)$$ where $G$ is an operator acting on an $L^p$ space of random vectors. By solving for this fixed point, the entire solution process $Y_t$ can be recovered, typically via conditional expectations. This shift effectively "eliminates" the time dimension from the core existence proof, significantly streamlining the analysis.

Methodological Tools

By reframing the problem in $L^p(\mathcal{F}_T)$, the authors can deploy standard functional analysis tools:

  1. Banach Contraction Mapping: For equations where the driver is Lipschitz continuous, Banach’s theorem provides a straightforward path to proving existence and uniqueness. This approach is robust enough to handle multidimensional cases and equations with jumps.
  2. Schauder and Krasnoselskii Theorems: For non-Lipschitz equations—such as those with superlinear or quadratic growth—the authors employ Schauder-type arguments. This requires proving that the mapping $G$ is compact.
  3. Infinite-Dimensional Sobolev Spaces: To establish the necessary compactness in a stochastic setting, the authors utilize Sobolev spaces corresponding to infinite-dimensional Gaussian measures. This allows them to find compact sets within the $L^p$ spaces, ensuring that the mathematical "shape" of the problem permits a solution.

Key Insights and Applications

The BSE framework is remarkably inclusive, covering classical BSDEs, semimartingale backward equations, and the semimartingale Bellman equations used in optimal control.

  • Path-Dependency: The authors provide concrete conditions under which BSDEs with functional, path-dependent drivers have unique solutions. They utilize a "path-patching" technique, showing that solutions can be joined together over small time intervals where the mapping is a contraction.
  • Mean-Field and McKean-Vlasov Equations: The paper derives existence results for multidimensional mean-field BSDEs with drivers of quadratic growth. This is a significant advancement, as previous work was often limited to one-dimensional cases or required very small terminal conditions.
  • Superlinear Growth: Multidimensional BSDEs with quadratic growth are notoriously difficult. By leveraging fixed point approach and compactness arguments, the authors offer new existence results for these challenging cases.

Conclusion and Future Directions

This research establishes a rigorous foundation for Lipschitz BSEs in $L^p$ spaces for any $p \in (1, \infty]$. It demonstrates that several landmark results in the field (e.g., those by Pardoux, Peng, and Delong) are specific cases of this broader BSE theory.

However, some challenges remain. The results for non-Lipschitz equations rely heavily on compactness assumptions and Sobolev spaces, which may be restrictive for certain stochastic processes. Additionally, in some path-dependent scenarios, the theory still requires either a small Lipschitz constant or a short time horizon. Future work may focus on relaxing these "smallness" requirements and developing numerical algorithms based on these fixed point iterations to compute solutions for practical engineering and financial applications.

Summary updated

Abstract

In this paper, we introduce a class of backward stochastic equations (BSEs) that extend classical BSDEs and include many interesting examples of generalized BSDEs as well as semimartingale backward equations. We show that a BSE can be translated into a fixed point problem in a space of random vectors. This makes it possible to employ general fixed point arguments to find solutions. For instance, Banach's contraction mapping theorem can be used to derive general existence and uniqueness results for equations with Lipschitz coefficients, whereas Schauder-type fixed point arguments can be applied to non-Lipschitz equations. The approach works equally well for multidimensional as for one-dimensional equations and leads to results in several interesting cases such as equations with path-dependent coefficients, anticipating equations, McKean-Vlasov type equations and equations with coefficients of superlinear growth.
BibTeX
@article{cheridito2017bse,
  author = {Patrick Cheridito and Kihun Nam},
  title = {{BSE's, BSDE's and fixed point problems}},
  journal = {The Annals of Probability},
  year = {2017},
  doi = {10.1214/16-aop1149},
}

Loading related content...

Figure