Metraipolis › archive › dc_5c5e200feffb
Generalization Bounds for Flow-matching Generative Models for Intrinsically Low-dimensional Data — arXiv:2610.02663
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.02663
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T07:57:57.919Z
- SHA-256 of the archived text
551cf128ac87cec635feaa162bfe207a0a77e9650a96b6d81deeb215996fee9c- Stable URL of this page
https://metraipolis.com/documento/dc_5c5e200feffb
How to cite
Generalization Bounds for Flow-matching Generative Models for Intrinsically Low-dimensional Data — arXiv:2610.02663 — arXiv (en), https://arxiv.org/abs/2610.02663 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5c5e200feffb (Metraipolis archive dc_5c5e200feffb, sha256 551cf128ac87cec635feaa162bfe207a0a77e9650a96b6d81deeb215996fee9c)
BibTeX
@misc{dc5c5e200feffb,
title = {Generalization Bounds for Flow-matching Generative Models for Intrinsically Low-dimensional Data — arXiv:2610.02663},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.02663},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5c5e200feffb; sha256 551cf128ac87cec635feaa162bfe207a0a77e9650a96b6d81deeb215996fee9c},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.02663. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Despite the remarkable empirical success of flow-matching models, their statistical generalization guarantees remain underdeveloped. Existing analyses often impose restrictive assumptions on the estimated velocity field and yield convergence rates that fail to reflect the intrinsic low-dimensional structure common in real data, such as natural images and molecular geometries. In this work, we study the statistical generalization of flow-matching models for learning an unknown distribution $P_{\mathrm{data}}$ from finitely many samples. We derive finite-sample error bounds on the learned generative distribution, measured in the Wasserstein-$p$ distance, for all $p\geq 1$. Specifically, given $n$ i.i.d. samples from $P_{\mathrm{data}}$, we show that, for every $d>d_p^\ast(P_{\mathrm{data}})$ and appropriately chosen network architectures and hyperparameters, the learned distribution $\widehat{P}^{\mathrm{FM}}$ satisfies $ \mathbb{W}_p(\widehat{P}^{\mathrm{FM}},P_{\mathrm{data}}) \lesssim n^{-1/d}+n^{-1/(2p)}\bigl(\log(1/\xi)\bigr)^{1/(2p)}$ with probability at least $1-\xi$, where $d_p^\ast(P_{\mathrm{data}})$ denotes the Wasserstein-$p$ dimension of the target measure. Our results demonstrate that flow matching naturally adapts to the intrinsic geometry of data and mitigates the curse of dimensionality, as the convergence exponent depends on the intrinsic rather than ambient dimension. These guarantees remain meaningful in high-dimensional regimes and provide a theoretical explanation for the empirical success of flow matching on structured data distributions under substantially more relaxed assumptions than those in existing analyses.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-05 in cs.AI).
Categorie arXiv: Machine Learning (stat.ML) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistics Theory (math.ST).
Pagina del lavoro: https://arxiv.org/abs/2610.02663 — 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 2026-10-06T07:47:59.577Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).