Robust Filtering of L\'evy-driven Stochastic Models

Harsha Honnappa; Sharan Srinivasan; Vijay Gupta

arxiv: 2602.14310 · v2 · submitted 2026-02-15 · 🧮 math.PR

Robust Filtering of L\'evy-driven Stochastic Models

Sharan Srinivasan , Vijay Gupta , Harsha Honnappa This is my paper

Pith reviewed 2026-05-15 21:40 UTC · model grok-4.3

classification 🧮 math.PR

keywords robust filteringLévy processesrough pathscadlag pathsnonlinear filteringStratonovich decompositionMarcus decompositionjump processes

0 comments

The pith

The robust nonlinear filter for Lévy-driven stochastic models is continuous in rough p-variation topologies on cadlag path space.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that a nonlinear filter for stochastic models driven by Lévy processes, where signal and observation share Brownian and jump noise, depends continuously on the observation path. This property is crucial because real observations often deviate from the model, for example when reconstructed from discrete samples. For finite-jump processes, continuity holds in both rough p-variation and p-variation topologies without separability conditions on jumps. For infinite-jump processes, continuity is shown in a modified rough p-variation topology under a separability assumption. The proofs rely on Stratonovich and Marcus flow decompositions to obtain geometric rough path lifts that provide pathwise guarantees.

Core claim

We construct a version of the filter and establish its continuity in two regimes. For processes with finitely many jumps on compact intervals, we prove continuity in both the rough p-variation and p-variation topologies on cadlag path space, without requiring a separability condition on the jump coefficients. For processes with infinitely many jumps, we prove continuity in a modified rough p-variation topology adapted to cadlag geometric rough paths, under an additional separability assumption. In both cases, our approach relies on Stratonovich and Marcus flow decompositions rather than the Itô-based methods of recent work. The resulting geometric rough-path lifts yield pathwise convergence,

What carries the argument

Stratonovich and Marcus flow decompositions yielding geometric rough-path lifts of the coupled system on cadlag space.

If this is right

The filter admits pathwise convergence when the observation path is approximated in the p-variation topology.
Discrete-time samples can be used to construct the rough path lift and evaluate the filter without knowledge of the probability law.
Continuity extends to both finite and infinite activity Lévy processes under the stated conditions.
The method avoids measure-dependent Itô techniques in favor of pathwise geometric constructions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The framework may extend to other jump-driven rough differential equations beyond filtering.
Future work could investigate whether the separability assumption for infinite jumps can be removed by adjusting the topology.
Numerical algorithms based on these lifts could enable robust filtering in applications with sampled jumpy data such as financial time series.

Load-bearing premise

The separability assumption on the jump coefficients is required to prove continuity for processes with infinitely many jumps.

What would settle it

A concrete counterexample where the filter map is discontinuous for an infinite-activity Lévy process whose jump coefficients violate the separability assumption.

read the original abstract

We study robust nonlinear filtering for stochastic models driven by L\'evy processes, where the signal and observation processes are coupled through common Brownian and jump noise. Robustness, defined as the continuous dependence of the filter on the observation path, is essential whenever the observation process deviates from the idealized model, for instance when a path must be reconstructed from discrete-time samples. This question is well understood for continuous semimartingale systems but largely open in the presence of jumps. We construct a version of the filter and establish its continuity in two regimes. For processes with finitely many jumps on compact intervals, we prove continuity in both the rough $p$-variation and $p$-variation topologies on cadlag path space, without requiring a separability condition on the jump coefficients. For processes with infinitely many jumps, we prove continuity in a modified rough $p$-variation topology adapted to cadlag geometric rough paths, under an additional separability assumption. In both cases, our approach relies on Stratonovich and Marcus flow decompositions rather than the It\^o-based methods of recent work. The resulting geometric rough-path lifts yield pathwise convergence guarantees and can be constructed directly from discrete observations without knowledge of the underlying probability law.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives pathwise continuity results for robust nonlinear filters on Lévy-driven systems with jumps via rough-path lifts, but the infinite-activity case hinges on an uncharacterized separability assumption.

read the letter

The core advance here is extending robust filtering continuity from continuous semimartingales to Lévy processes that include jumps. They build a filter version and show it depends continuously on the observation path in rough p-variation and p-variation topologies on cadlag space. For finite jumps on compact intervals this holds without extra conditions on the coefficients. For infinite jumps they use a modified rough topology and Marcus flow decompositions, but only under a separability assumption on the jump coefficients. The shift away from Itô methods toward Stratonovich and Marcus lifts is the technical move that lets them get pathwise guarantees directly from discrete samples without knowing the law in advance. That part is clean and addresses a real gap for applications in finance and signal processing where jumps matter. The finite-jump regime looks solid on the evidence in the abstract. The infinite-jump part is the weaker link because the separability condition is stated but not illustrated with concrete Lévy measures or counterexamples, so it is hard to judge how often it holds for typical infinite-activity processes. No circularity or invented entities show up. This is technical work aimed at specialists in rough paths and nonlinear filtering. A reader already comfortable with geometric rough paths and cadlag lifts will get the most out of it. The paper is coherent enough on its own terms to deserve a serious referee rather than a desk reject, though the referee will likely ask for more on the scope of the separability assumption.

Referee Report

2 major / 2 minor

Summary. The paper constructs a version of the nonlinear filter for Lévy-driven signal-observation systems coupled by common Brownian and jump noise, and proves its continuity with respect to the observation path in two regimes. For finite-activity jump processes on compact intervals, continuity holds in both the rough p-variation and p-variation topologies on càdlàg path space without a separability condition on the jump coefficients. For infinite-activity processes, continuity is established in a modified rough p-variation topology adapted to càdlàg geometric rough paths, under an additional separability assumption on the jump coefficients. The proofs rely on Stratonovich and Marcus flow decompositions rather than Itô methods, yielding pathwise convergence that can be constructed directly from discrete observations.

Significance. If the continuity claims hold, the work extends robust filtering from continuous semimartingales to Lévy-driven models with jumps, supplying pathwise guarantees that are directly applicable to filter reconstruction from discrete samples. The geometric rough-path approach via flow decompositions provides a clean framework that avoids measure-theoretic issues and may generalize to other jump-driven systems.

major comments (2)

[Section 4] Section 4 (infinite-activity regime): The separability assumption on the jump coefficients is load-bearing for the modified rough p-variation topology and the associated Marcus flow decomposition, yet the manuscript provides neither concrete examples of Lévy measures satisfying the condition nor counterexamples showing processes that fail it. Without such characterization (e.g., bounds on the Lévy measure or explicit verification for standard infinite-activity processes such as variance-gamma or stable processes), the scope of the continuity result remains unclear.
[Theorem 3.1] Theorem 3.1 and the finite-jump continuity statement: The claim of continuity in both rough p-variation and p-variation topologies is stated without an explicit comparison of the two topologies or a discussion of when one is strictly stronger; this weakens the assertion that the result holds 'in both' topologies simultaneously.

minor comments (2)

[Abstract] The abstract and introduction use 'cadlag path space' without a precise reference to the underlying metric or topology; a short clarification of the Skorokhod-type metric employed would improve readability.
[Section 4.1] Notation for the modified rough p-variation metric in the infinite-jump case is introduced without an explicit formula or comparison to the standard rough p-variation distance; adding the definition in the main text rather than an appendix would aid verification.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading and constructive comments. We address each major point below and will revise the manuscript accordingly to improve clarity.

read point-by-point responses

Referee: [Section 4] Section 4 (infinite-activity regime): The separability assumption on the jump coefficients is load-bearing for the modified rough p-variation topology and the associated Marcus flow decomposition, yet the manuscript provides neither concrete examples of Lévy measures satisfying the condition nor counterexamples showing processes that fail it. Without such characterization (e.g., bounds on the Lévy measure or explicit verification for standard infinite-activity processes such as variance-gamma or stable processes), the scope of the continuity result remains unclear.

Authors: We agree that explicit examples and a brief characterization would better delineate the scope of the separability assumption in the infinite-activity case. In the revised manuscript we will add a new remark in Section 4 that verifies the condition for standard processes (including variance-gamma and stable Lévy processes with index >1) and supplies a simple counter-example of a Lévy measure for which the coefficients fail to be separable. This addition will not alter the statements of the theorems but will make their applicability transparent. revision: yes
Referee: [Theorem 3.1] Theorem 3.1 and the finite-jump continuity statement: The claim of continuity in both rough p-variation and p-variation topologies is stated without an explicit comparison of the two topologies or a discussion of when one is strictly stronger; this weakens the assertion that the result holds 'in both' topologies simultaneously.

Authors: We thank the referee for this observation. Continuity in the rough p-variation topology is the stronger statement and automatically yields continuity in the p-variation topology; the converse does not hold in general. In the revision we will insert a short comparison paragraph immediately after the statement of Theorem 3.1 (and a corresponding sentence in the introduction) that recalls the relationship between the two topologies on càdlàg paths and explains why the proof yields the stronger result. This clarifies the simultaneous validity without changing any claims. revision: yes

Circularity Check

0 steps flagged

No circularity: claims rest on external rough-path theory and explicit flow decompositions

full rationale

The paper's central results establish continuity of a constructed filter in rough p-variation topologies for finite-jump Lévy processes without separability and in a modified topology for infinite-jump cases under an explicitly stated additional separability assumption. These proofs rely on Stratonovich/Marcus decompositions and geometric rough-path lifts drawn from prior literature, with no parameter fitting, self-definitional reductions, or load-bearing self-citations that collapse the derivation to its own inputs. The separability condition is introduced as an extra hypothesis rather than derived from the result itself, and the finite-jump regime avoids it entirely. No equations or steps reduce by construction to fitted quantities or renamed assumptions; the argument remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The work rests on standard results from rough path theory and stochastic analysis for Lévy processes; no free parameters or invented entities are introduced in the abstract.

axioms (2)

domain assumption Existence and properties of Stratonovich and Marcus flow decompositions for Lévy-driven systems
Invoked to obtain the geometric rough-path lifts that deliver pathwise convergence.
standard math Cadlag path space admits rough p-variation and modified rough p-variation topologies
Background topology used to state the continuity results.

pith-pipeline@v0.9.0 · 5514 in / 1385 out tokens · 44725 ms · 2026-05-15T21:40:31.435269+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

We construct a version of the filter and establish its continuity in two regimes... under an additional separability assumption... relies on Stratonovich and Marcus flow decompositions
IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Marcus lift... log-linear path function... βp topology adapted to cadlag geometric rough paths

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

The Variational Approach in Filtering and Correlated Noise
math.PR 2026-04 unverdicted novelty 7.0

A conditional variational principle for nonlinear filtering that handles correlated signal and observation noise, generalizing the Mitter-Newton formulation.