Lectures on stochastic sewing with applications

Basic Information

• Paper ID: 2510.12165
• Title: Lectures on stochastic sewing with applications
• Author: Oleg Butkovsky (Weierstrass Institute, Humboldt University Berlin, Simons Laufer Mathematical Sciences Institute)
• Classification: math.PR
• Publication Date: October 15, 2025
• Paper Link: https://arxiv.org/abs/2510.12165

Abstract

This is a set of lecture notes on stochastic sewing techniques, delivered in spring and summer 2025 at the University of Edinburgh and Beijing Institute of Technology. The objective is to introduce readers to stochastic sewing techniques and demonstrate how to successfully apply them to various problems in stochastic analysis, including: noise regularization of stochastic differential equations driven by Brownian motion or fractional Brownian motion, well-posedness of stochastic partial differential equations with irregular drift, investigation of averaging operators and local times, and analysis of numerical algorithms.

Research Background and Motivation

Core Problem

The core problem addressed in this paper is how to analyze bounds on integrals of the form: $\int f(\text{noise}(t) + \text{perturbation}(t)) dt, \quad \int f(\text{noise}(t) + \text{perturbation}(t)) d\text{noise}(t)$

where $f$ is a measurable, non-smooth function or even a Schwartz distribution, "noise" is a highly irregular process with known distribution, and "perturbation" is a process with unknown distribution but more regular than the noise.

Problem Significance

1. Noise regularization phenomenon: Deterministic systems may be ill-posed, but may become well-posed after adding random perturbations
2. Limitations of classical methods: The Veretennikov-Zvonkin transformation method heavily relies on good Itô formulas, making it difficult to apply to fractional Brownian motion or stochastic partial differential equations
3. Broad applicability: Stochastic sewing does not depend on Markov properties or semimartingale properties, with wider applicability

Research Motivation

Traditional Krylov-type bound estimation methods have limitations when dealing with stochastic differential equations driven by irregular noise. Stochastic sewing techniques, as an effective alternative to the Veretennikov-Zvonkin transformation method, provide a more general analytical framework.

Core Contributions

1. Systematic introduction: First systematic introduction of stochastic sewing techniques and their extensions
2. Unified framework: Incorporates seemingly different problems (well-posedness, numerical methods, regularization) into a unified stochastic sewing framework
3. Technical innovations:
   • Generalized coupling techniques for weak uniqueness
   • Shifted stochastic sewing lemma
   • Application of John-Nirenberg inequalities
   • Introduction of conditional Hölder norms
4. Application extensions: Demonstrates complete theoretical development from Brownian motion to Lévy processes, from strong well-posedness to weak well-posedness

Detailed Methodology

Core Technique: Stochastic Sewing Lemma

Classical sewing lemma (Feyel-De La Pradelle, Gubinelli) requires: $\|\delta A_{s,u,t}\|_E \leq \Gamma|t-s|^\gamma, \quad \gamma > 1$

Key improvement of stochastic sewing lemma (Lê): $\|\delta A_{s,u,t}\|_{L^m(\Omega)} \leq \Gamma_1|t-s|^{\gamma_1}, \quad \gamma_1 > \frac{1}{2}$$\|E_s\delta A_{s,u,t}\|_{L^m(\Omega)} \leq \Gamma_2|t-s|^{\gamma_2}, \quad \gamma_2 > 1$

where $\delta A_{s,u,t} := A_{s,t} - A_{s,u} - A_{u,t}$.

Key Innovations

1. Advantages of Conditional Expectation

The conditional expectation $E_s\delta A_{s,u,t}$ is typically much smaller than $\delta A_{s,u,t}$, which is the key advantage of stochastic sewing over deterministic sewing.

2. Shifted Stochastic Sewing Lemma

To address integral divergence issues, a modified simplex is introduced: $\Delta^{\text{shift}}_{[S,T]} := \{(s,t) \in [S,T]^2 : s < t, s-(t-s) \geq S\}$

The key condition becomes: $\|E_{s-(t-s)}\delta A_{s,u,t}\|_{L^m(\Omega)} \leq \Gamma_2|t-s|^{\gamma_2}$

3. Conditional Hölder Norm

A new seminorm is defined: $[f]_{C^\tau L^m([S,T])} := \sup_{(s,t)\in\Delta_{[S,T]}} \frac{\|f(t) - E^s f(t)\|_{L^m(\Omega)}}{|t-s|^\tau}$

This is weaker than the traditional Hölder seminorm but better suited to the stochastic setting.

Main Applications

1. Strong Uniqueness for Brownian Motion-Driven SDEs

Theorem 1.1 (Veretennikov-Zvonkin): Let $b: \mathbb{R}^d \to \mathbb{R}^d$ be a bounded measurable function. Then the SDE $dX_t = b(X_t)dt + dW_t, \quad X_0 = x$ has a unique strong solution.

Proof strategy:

1. Establish integral bounds: for $\gamma \in (0,1]$, $\left\|\int_S^T (b(W_r + x) - b(W_r + y))dr\right\|_{L^m(\Omega)} \leq C\|b\|_{C^\gamma}|x-y||T-S|^{\frac{1}{2}+\frac{\gamma}{2}}$
2. Extension to random perturbations: $\left\|\int_S^T (b(W_r + \phi_r) - b(W_r + \psi_r))dr\right\|_{L^m(\Omega)} \leq C(T-S)^{\frac{1}{2}+\frac{\gamma}{2}}(\text{regularity terms})$

2. Weak Uniqueness with Distributional Drift

Theorem 1.2: Let $b \in C^\gamma(\mathbb{R}^d, \mathbb{R}^d)$, $\gamma > -\frac{1}{2}$. Then the SDE has a weak solution and is weakly unique in the class $V(1+\frac{\gamma}{2})$.

Proof strategy:

• Use generalized coupling method
• Introduce auxiliary process $\tilde{X}^n$ satisfying: $d\tilde{X}^n_t = b_n(\tilde{X}^n_t)dt + \lambda(X_t - \tilde{X}^n_t)dt + dW_t$
• Balance parameter $\lambda$ to control two error terms

3. Convergence Rates of Numerical Methods

Theorem 1.4: For the Euler scheme, the convergence rate is: $\|\sup_{t\in[0,1]} |X_t - X^n_t|\|_{L^m(\Omega)} \leq Cn^{-\frac{1}{2}-\frac{\gamma}{2}+\varepsilon}$

Theorem 1.5: For SDEs driven by $\alpha$-stable processes: $\|\sup_{t\in[0,1]} |X_t - X^n_t|\|_{L^m(\Omega)} \leq Cn^{-\frac{1}{2}-\frac{\gamma}{\alpha}+\varepsilon}$

4. Technical Tools

John-Nirenberg Inequality

Theorem 5.6: If $E^s|A_t - A_s| \leq \Gamma$ a.s., then $\|\sup_{r\in[s,T]} |A_r - A_s|\|_{L^m(\Omega)} \leq Cm\Gamma$

This tool allows extending $L^2$ bounds to arbitrary $L^m$ bounds without loss of convergence rate.

Experimental Setup and Results

Theoretical Verification

The paper is primarily theoretical, with validity of the methods verified through:

1. New proofs of classical results: Reproof of the Veretennikov-Zvonkin theorem using stochastic sewing
2. Extension to new settings: Successful treatment of cases driven by fractional Brownian motion, Lévy processes, etc.
3. Optimal convergence rates: Convergence rates of numerical methods are sharp in the sense of known optimality

Key Estimate Verification

For the Brownian motion case, the key heat kernel bound: $\|P_t f\|_{C^1} \leq Ct^{\frac{\gamma-1}{2}}\|f\|_{C^\gamma}$

For $\alpha$-stable processes: $\|P^\alpha_t f\|_{C^1} \leq Ct^{\frac{\gamma-1}{\alpha}}\|f\|_{C^\gamma}$

Related Work

Classical Methods

1. Veretennikov-Zvonkin transformation: Relies on Itô formulas and PDE techniques
2. Girsanov theorem method: Requires square-integrability of drift
3. Deterministic sewing lemma: Requires regularity $\gamma > 1$

Development of Stochastic Sewing

1. Lê (2018): Original stochastic sewing lemma
2. Gerencsér (2022): Shifted stochastic sewing
3. This work: Systematized theory and broad applications

Application Domain Extensions

• Rough path theory
• Stochastic partial differential equations
• McKean-Vlasov equations
• Fast-slow systems

Conclusions and Discussion

Main Conclusions

1. Unified framework: Stochastic sewing provides a unified method for handling SDEs with irregular driving processes
2. Technical flexibility: The form of stochastic sewing can be adjusted and modified according to specific problems
3. Broad applicability: Complete theory from Brownian motion to Lévy processes, from strong to weak well-posedness

Limitations

1. Technical complexity: Requires developing specialized technical tools for different problems
2. Regularity requirements: Still requires certain regularity assumptions
3. Computational complexity: Practical applications may involve complex computations

Future Directions

1. More general driving processes: Extension to broader classes of stochastic processes
2. Multi-parameter time: Development of multi-parameter stochastic sewing theory
3. Numerical algorithm optimization: Design of more efficient numerical methods based on stochastic sewing theory

In-Depth Evaluation

Strengths

1. Theoretical innovation: The stochastic sewing lemma is an important generalization of the classical sewing lemma
2. Strong systematization: First systematic organization of theory and applications in this field
3. Technical depth: Demonstrates how to adjust technical tools for specific problems
4. Broad applications: Covers multiple important problems including well-posedness, numerical methods, and regularization

Weaknesses

1. Primarily expository: Most results are from previously published work
2. Lack of numerical experiments: Theoretical results lack numerical verification
3. High technical threshold: Requires strong background in stochastic analysis

Impact

1. Educational value: As lecture notes, provides important learning resources for the field
2. Research guidance: Provides clear technical roadmap for related research
3. Method promotion: Demonstrates application potential of stochastic sewing in different problems

Applicable Scenarios

1. Theoretical research: Well-posedness studies of stochastic differential equations and stochastic partial differential equations
2. Numerical analysis: Analysis of numerical methods for SDEs with irregular coefficients
3. Applied mathematics: Stochastic modeling in financial mathematics, biomathematics, and other fields

References

The paper includes important literature in the field, particularly:

• Lê20 K. Lê. A stochastic sewing lemma and applications
• Gub04 M. Gubinelli. Controlling rough paths
• Ver80, Zvo74 Classical Veretennikov-Zvonkin theory
• BDG21, BDG22 The author's work on numerical methods

These lecture notes provide a comprehensive and in-depth introduction to stochastic sewing theory and serve as an important reference for researchers and learners in the field. While primarily expository in nature, its systematic organization and clear technical roadmap demonstrate significant academic value.