TL;DR
This paper introduces a new notion of capacity on abstract Wiener spaces, enabling a more robust calculus for functions of bounded variation (BV). A key result is that the total variation measure of BV functions is absolutely continuous with respect to this new capacity, allowing for extensions of standard calculus rules. As a significant application, the authors demonstrate that the maximum of a Wiener process, a seemingly simple object, possesses a second derivative that is a measure singular with respect to the Wiener measure itself.
What Problem Does This Paper Address?
The paper tackles the challenge of extending calculus rules, particularly for functions of bounded variation (BV), to the infinite-dimensional setting of abstract Wiener spaces. In finite dimensions, BV functions are well-understood, and their properties are crucial in areas like image processing and geometric measure theory. However, extending these concepts to infinite dimensions, where the underlying space is a Wiener space (a space of continuous paths), is significantly more complex. Standard tools like Malliavin calculus, while powerful, often struggle with functions that are not smooth enough. Specifically, the paper aims to:
- Develop a BV calculus: Extend the rules of calculus (like integration by parts and Leibniz rule) to BV functions on Wiener spaces, which are not necessarily smooth. This requires overcoming the limitations of existing tools.
- Characterize regularity of stochastic processes: Investigate the regularity of specific stochastic processes, particularly the maximum of a Wiener process, which is known to be only once differentiable in a standard sense. The paper seeks to understand its higher-order derivatives as measures.
Methods
The core of the paper lies in the introduction of a novel BV-capacity on abstract Wiener spaces. This capacity is defined in a way that refines existing notions and is designed to be compatible with BV functions. The authors prove a crucial result: the total variation measure of any BV function is absolutely continuous with respect to this BV-capacity. This absolute continuity is the key technical tool that allows them to extend calculus rules.
Key techniques include:
- BV-capacities: A new definition of capacity tailored for BV functions on Wiener spaces.
- Absolute Continuity: Proving that the total variation measure of a BV function is absolutely continuous with respect to the defined BV-capacity. This is a generalization of the concept of absolute continuity for measures.
- Malliavin Calculus: Leveraging existing tools from Malliavin calculus for smooth functions and extending them to the BV setting.
- Approximation Techniques: Using sequences of smooth functions to approximate BV functions and transfer properties from the smooth to the BV realm.
- Integration by Parts and Leibniz Rule Extensions: Rigorously extending these fundamental calculus rules to BV functions under suitable conditions.
- Precise Representatives: Utilizing the concept of precise representatives for functions to handle measurability and convergence issues in the BV setting.
Datasets
Not discussed in the paper. The paper deals with theoretical concepts on abstract Wiener spaces and the classical Wiener space, rather than empirical datasets.
Key Insights and Ideas
- A New BV-Capacity is Key: The introduction of a BV-capacity is the central innovation. Its absolute continuity with respect to the total variation measure of BV functions is the linchpin for extending calculus rules.
- Bridging Smoothness Gaps: The BV-capacity acts as a bridge, allowing calculus rules to be applied to functions that are not smooth enough for standard Malliavin calculus.
- The Maximum of a Wiener Process is "More Singular" Than Expected: A surprising result is that the maximum of a Wiener process, while only once differentiable, has a second derivative that is a measure. Crucially, this measure is singular with respect to the underlying Wiener measure. This implies that the "jumps" or "singularities" of this second derivative do not align with the typical behavior of Wiener paths.
- Concentration of the Second Derivative: The paper shows that the measure representing the second derivative of the Wiener process maximum is concentrated on trajectories that attain the global maximum exactly twice. This provides a precise geometric interpretation of the singularities.
Key Results
- Absolute Continuity Theorem: For any BV function $u$ on an abstract Wiener space, its total variation measure $|Du|$ is absolutely continuous with respect to a newly defined BV-capacity $C$.
- Extended Calculus Rules: The paper provides rigorous extensions of the integration by parts formula and the Leibniz rule for products of BV functions and differentiable functions.
- Second Derivative of Wiener Maximum: The maximum of a Wiener process $M = \sup_{0 \le t \le T} W_t$ admits a second derivative $D^2M$ as a measure on the Wiener space.
- Singularity of $D^2M$: The total variation measure of $D^2M$ is singular with respect to the Wiener measure $P$.
- Concentration of $D^2M$: The measure $D^2M$ is concentrated on paths that attain the global maximum exactly twice.
Limitations and Future Work
- Coincidence of Capacities: The authors note that they do not know whether their BV-capacity ($C_{BV}$) always coincides with the cylindrical capacity ($C_{cyl}$). This is an open question.
- General Chain Rule for BV Functions: The paper explicitly states that it does not attempt to prove a general chain rule for BV functions, as this would require a deep investigation into the structure of perimeter sets. This remains an area for future research.
- Connection to Hausdorff Measures: The authors mention that exploring connections between BV-capacities and 1-codimensional Hausdorff measures is a direction for future investigation.
- Uniqueness of Minimum for BV-Capacity: While the infimum in the definition of BV-capacity is attained, the uniqueness of the minimum is not clear, unlike in the case of $L^p$ norms. This is a point of difference and potential future study.
Summary updated