Metraipolis › archive › dc_bc3f47cde643

Towards Hyperparameter Transfer for Differentially Private Optimization — arXiv:2610.22976

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.22976
License (as archived): CC0-1.0
Archived: 2026-10-19T08:46:42.015Z
SHA-256 of the archived text: 3551d401deebc69a25d4d1a0338c56c158db2965cd1bd56d096466da95d61cfd
Stable URL of this page: https://metraipolis.com/documento/dc_bc3f47cde643

How to cite

Towards Hyperparameter Transfer for Differentially Private Optimization — arXiv:2610.22976 — arXiv (en), https://arxiv.org/abs/2610.22976 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bc3f47cde643 (Metraipolis archive dc_bc3f47cde643, sha256 3551d401deebc69a25d4d1a0338c56c158db2965cd1bd56d096466da95d61cfd)

BibTeX

@misc{dcbc3f47cde643,
  title = {Towards Hyperparameter Transfer for Differentially Private Optimization — arXiv:2610.22976},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.22976},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bc3f47cde643; sha256 3551d401deebc69a25d4d1a0338c56c158db2965cd1bd56d096466da95d61cfd},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22976. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Scaling laws and hyperparameter tuning remain central challenges in training large-scale models, particularly under differential privacy (DP), where optimization is further complicated by noise injection and per-sample gradient clipping. While Maximal Update Parametrization ($\mu$P) allows for zero-shot hyperparameter transfer in standard training, we show that it fails under DP constraints due to the distinct spectral properties of high-dimensional noise versus low-rank gradients. In this work, we introduce DP-$\mu$P, a theoretically grounded framework that extends $\mu$P to the privacy-preserving setting. By analyzing the interaction between gradient clipping, noise injection, and feature learning, we derive a Signal-to-Noise Ratio (SNR) condition that governs layer-wise learning rate scaling. Combining this with our proposed spectral scaling formulation, we provide specific scaling rules for DP-SGD, DP-Adam, and DP-Muon optimizers. Empirically, we demonstrate that DP-$\mu$P enables effective zero-shot transfer of learning rates across model widths on diverse architectures (MLP, ViT, GPT-2), matching or exceeding the performance of individually tuned baselines while significantly reducing computational costs.

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/2610.22976 — 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 2026-10-19T08:41:54.790Z dalla pagina https://arxiv.org/list/cs.CR/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).