Metraipolis › archive › dc_4fb03429be0a

Adaptive Bregman Proximal Stochastic Gradient with a Stabilized Barzilai--Borwein Step Size — arXiv:2608.12009

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.12009
License (as archived)
CC0-1.0
Archived
2026-10-19T05:40:21.921Z
SHA-256 of the archived text
1350e145659e30de78d87aa9562408226e0ccdf4f5bd4bd7da513877f6d2b8f9
Stable URL of this page
https://metraipolis.com/documento/dc_4fb03429be0a

How to cite

Adaptive Bregman Proximal Stochastic Gradient with a Stabilized Barzilai--Borwein Step Size — arXiv:2608.12009 — arXiv (en), https://arxiv.org/abs/2608.12009 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4fb03429be0a (Metraipolis archive dc_4fb03429be0a, sha256 1350e145659e30de78d87aa9562408226e0ccdf4f5bd4bd7da513877f6d2b8f9)
BibTeX
@misc{dc4fb03429be0a,
  title = {Adaptive Bregman Proximal Stochastic Gradient with a Stabilized Barzilai--Borwein Step Size — arXiv:2608.12009},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.12009},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4fb03429be0a; sha256 1350e145659e30de78d87aa9562408226e0ccdf4f5bd4bd7da513877f6d2b8f9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.12009. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Bregman proximal stochastic gradient (BPSG) methods bring variance-reduced composite optimization to objectives whose geometry is poorly captured by Euclidean smoothness. Their performance, however, remains sensitive to the step size: raw stochastic curvature estimates can fluctuate sharply, whereas line searches add repeated proximal evaluations. We introduce Ada-BPSG, a line-search-free BPSG method that couples the SAGA gradient table with a stabilized Barzilai--Borwein (BB) candidate. A mediant aggregates incremental secant information so that nearly singular local ratios receive little weight, and an explicit safeguard translates the resulting curvature estimate into the bounded step-size sequence required for convergence. This design yields a direct analytical chain from relative smoothness and component-wise variance control to convergence in finite-dimensional normed spaces. We prove an $O(n/K)$ ergodic rate for convex objectives, a restarted linear rate under relative quadratic growth, and an $O(1/K)$ bound for a Bregman proximal residual in the nonconvex setting. On logistic regression and sparse nonnegative matrix factorization, Ada-BPSG combines low objective values with substantially less sensitivity to the initial step size than standard variance-reduced baselines, while avoiding line search.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-13 in cs.LG).
Categorie arXiv: Optimization and Control (math.OC) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.12009 — 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 2026-10-19T05:18:32.552Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-13?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).