Metraipolis › archive › dc_9a4c02009986

Differential Privacy of Gradient Descent on Perturbed Objectives — arXiv:2610.02716

Source
arXiv
Original
https://arxiv.org/abs/2610.02716
License (as archived)
CC0-1.0
Archived
2026-10-05T23:06:29.720Z
SHA-256 of the archived text
03f89b0d8c34d9d4bf2fe67dda95ff009e3c12c454f6f634863c30804defe271
Stable URL of this page
https://metraipolis.com/documento/dc_9a4c02009986

How to cite

Differential Privacy of Gradient Descent on Perturbed Objectives — arXiv:2610.02716 — arXiv, https://arxiv.org/abs/2610.02716 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9a4c02009986 (Metraipolis archive dc_9a4c02009986, sha256 03f89b0d8c34d9d4bf2fe67dda95ff009e3c12c454f6f634863c30804defe271)
BibTeX
@misc{dc9a4c02009986,
  title = {Differential Privacy of Gradient Descent on Perturbed Objectives — arXiv:2610.02716},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.02716},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9a4c02009986; sha256 03f89b0d8c34d9d4bf2fe67dda95ff009e3c12c454f6f634863c30804defe271},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.02716. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Objective perturbation adds a random linear term to a regularized empirical risk and releases the exact perturbed minimizer. We study the finite computation obtained by releasing the $N$-th iterate of deterministic gradient descent on $w\mapsto F(w;S)+\langle z,w\rangle$, where $z\sim\mathcal N(0,\sigma^2I_d)$ is drawn once before optimization. For strongly convex and smooth objectives with Lipschitz Hessian, we prove an explicit condition under which the map $z\mapsto w_N$ is a $C^1$-diffeomorphism on the bounded domains used in the privacy argument, with a quantitative lower bound on the smallest singular value of its Jacobian. This permits a direct change-of-variables analysis of the finite iterate. For generalized linear models, the resulting privacy-profile bound has no explicit ambient-dimension factor once the iteration condition holds, and its finite-iteration correction decreases geometrically. By letting the free truncation parameter grow slowly with $N$, we recover the corresponding exact-minimizer certificate in the limit. We also bound the expected excess empirical risk by $d\sigma^2/(2\mu)$ plus a geometrically decreasing optimization term, and transfer the result to population risk without an additional multiplicative condition-number factor in the leading statistical terms.

Tema degli esploratori: fiducia digitale e crittografia (elenco arXiv cs.CR, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Cryptography and Security (cs.CR); Machine Learning (stat.ML).
Pagina del lavoro: https://arxiv.org/abs/2610.02716 — 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 2026-10-05T23:01:08.556Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).