Metraipolis › archive › dc_166b102750f1

Descending Everest with First-Order Oracles: Oracle Complexity of Convex Optimization under $(L_0,L_1)$-Smoothness — arXiv:2610.21461

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.21461
License (as archived)
CC0-1.0
Archived
2026-10-16T10:28:36.393Z
SHA-256 of the archived text
5686ab3585ee1547988841a512def69f7384d6456406e70d38ca0541b94ff4b8
Stable URL of this page
https://metraipolis.com/documento/dc_166b102750f1

How to cite

Descending Everest with First-Order Oracles: Oracle Complexity of Convex Optimization under $(L_0,L_1)$-Smoothness — arXiv:2610.21461 — arXiv (en), https://arxiv.org/abs/2610.21461 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_166b102750f1 (Metraipolis archive dc_166b102750f1, sha256 5686ab3585ee1547988841a512def69f7384d6456406e70d38ca0541b94ff4b8)
BibTeX
@misc{dc166b102750f1,
  title = {Descending Everest with First-Order Oracles: Oracle Complexity of Convex Optimization under $(L_0,L_1)$-Smoothness — arXiv:2610.21461},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21461},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_166b102750f1; sha256 5686ab3585ee1547988841a512def69f7384d6456406e70d38ca0541b94ff4b8},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21461. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

For convex $(L_0,L_1)$-smooth functions, the curvature may grow together with the gradient norm. Therefore, even when $\|x^0-x^*\|\le R$, the gradient and the optimality gap at the starting point may be exponentially large in $L_1R$, and the bound $\sqrt{L_0R^2/\varepsilon}$, known for the accelerated method on functions with Lipschitz gradient, no longer describes the complexity of the problem. We determine the oracle complexity of this class for arbitrary deterministic first-order methods: for $L_0R^2/\varepsilon\ge1$, the optimal number of oracle calls $N$ satisfies $1+N=\widetilde\Theta(\sqrt{L_0R^2/\varepsilon}+L_1R)$. The upper bound reads $O(\sqrt{L_0R^2/\varepsilon}+(1+L_1R)\log(2+L_1R))$: the method first passes through the large-gradient regime in $O(1+L_1R)$ iterations with an inexact line search, and then the standard accelerated method with a constant stepsize takes over. The lower bound is obtained by means of chains and the orthogonal invariance principle on functions conjugate to strongly convex functions on a polytope. If the initial optimality gap is at most $4\varepsilon$, the optimal bound becomes $1+N=\widetilde\Theta(\sqrt{L_0R^2/\varepsilon}+(L_1R)^{2/3})$.

Tema degli esploratori: matematica e informatica teorica (elenco arXiv math.OC, nuovi lavori del giorno).
Categorie arXiv: Optimization and Control (math.OC).
Pagina del lavoro: https://arxiv.org/abs/2610.21461 — 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 2026-10-16T08:49:47.925Z dalla pagina https://arxiv.org/list/math.OC/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).