Metraipolis › archive › dc_713dcd855760
Weak and strong discretization errors for signatures of Gaussian processes — arXiv:2611.08588
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.08588
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-09T22:43:07.995Z
- SHA-256 of the archived text
9bc39078b30cb18a86a686da20951746b6482b9caca7b9920fd7da43eeb4e381- Stable URL of this page
https://metraipolis.com/documento/dc_713dcd855760
How to cite
Weak and strong discretization errors for signatures of Gaussian processes — arXiv:2611.08588 — arXiv (en), https://arxiv.org/abs/2611.08588 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_713dcd855760 (Metraipolis archive dc_713dcd855760, sha256 9bc39078b30cb18a86a686da20951746b6482b9caca7b9920fd7da43eeb4e381)
BibTeX
@misc{dc713dcd855760,
title = {Weak and strong discretization errors for signatures of Gaussian processes — arXiv:2611.08588},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.08588},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_713dcd855760; sha256 9bc39078b30cb18a86a686da20951746b6482b9caca7b9920fd7da43eeb4e381},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.08588. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Let $X$ be a continuous Gaussian process in $\mathbb{R}^d$, $d\ge2$, with independent components of roughness $H\in(\tfrac14,1)$, possibly with a linear drift, and let $X^n$ be its piecewise-linear interpolant on a mesh of width $\delta$. We study how well the truncated signature of $X^n$ approximates that of the natural lift of $X$, in mean (the weak error) and in $L^2$ (the strong error). Over a class of drivers defined by quantitative regularity conditions, the weak error at every fixed truncation level is $O(\delta^{\gamma(H)})$ with $\gamma(H)=\min(2H,4H-1)$, uniformly over the class, and for truncation levels $M\ge4$ fractional Brownian motion attains this rate. For centred drivers the weak error vanishes at levels $1$ to $3$, Gaussian or not. It first appears at level~$4$, where we compute its leading constant for independent fractional Brownian motion in $d=2$: on $(\tfrac12,1)$ it is $1/(2\sqrt3)$, independent of $H$, and on $(\tfrac14,\tfrac12)$ it is given by a $\zeta$-series. For fractional Brownian motion with drift, a level-$3$ error of the same order appears for $H\ge\tfrac12$. Below $\tfrac12$ the drift reaches the rate only at levels $\ge5$ (proved for $d\ge3$). The strong error is of exact order $\delta^{2H-1/2}$ for fractional Brownian motion with $H\in(\tfrac12,1)$, so in mean square the weak error lies a full power of $\delta$ below it. For $H\le\tfrac12$ the strong error is $O(\delta^{\vartheta})$ for every $\vartheta<2H-\tfrac12$, uniformly over the class, and the exponent $2H-\tfrac12$ cannot be improved. Finally, we construct pairs of laws in the class whose mesh data have the same law but whose expected signatures differ by order $\delta^{\gamma(H)}$. Hence no quantity computed from the mesh data has mean within $o(\delta^{\gamma(H)})$ of the expected signature uniformly over the class.
Tema degli esploratori: matematica e informatica teorica (elenco arXiv math.PR, nuovi lavori del giorno).
Categorie arXiv: Probability (math.PR).
Pagina del lavoro: https://arxiv.org/abs/2611.08588 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-09T22:36:02.232Z dalla pagina https://arxiv.org/list/math.PR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).