Dimension-Optimal Dynamic Regret in Exp-Concave Online Learning N280AVNNBY

URI	`m2://res/N280AVNNBY`
Kind	preprint
Creators	Yanyi Li, Wendao Wu, Haihan Zhang, Qinshi Wang, Chenheng Zhang, Chunyuan Zheng, Cong Fang, Haoxuan Li, Zhouchen Lin
Date	2026-10-14
Subjects	arxiv cs.LG, arxiv math.OC
Scope	border

Expressions

Date	2026-10-14
Ids	arXiv 2610.20187

The mentions of the inventory that name this resource. Each name links to the card of its inventory row.

Item	Kind	Where	Source
Polynomial-bit implementation	statement	theorem B.18	arxiv
1FYBKPM63N	statement	theorem B.16	arxiv
3NBHAPZ6RY	statement	theorem C.4	arxiv
Block-hypercube lower bound in effective dimension $m$	statement	lemma C.2	arxiv
Curved dynamic regret	statement	theorem 4.1	arxiv
Domination and surrogate exp-concavity	statement	lemma B.1	arxiv
677NPA1X5A	statement	corollary B.17	arxiv
Shifted drawdown energy	statement	lemma B.8	arxiv
One-step KL inequality	statement	lemma B.6	arxiv
Sequential two-point risk	statement	lemma C.1	arxiv
Prefix overhead to every surviving history	statement	lemma B.3	arxiv
Computational minimax closure, up to logarithms	statement	corollary 5.2	arxiv
Gaussian mixability gap	statement	lemma B.5	arxiv
Dynamic projected gradient descent	statement	lemma B.13	arxiv
Matching normalized lower frontier	statement	theorem 5.1	arxiv
Square-root KL movement	statement	lemma B.7	arxiv
Static lower bound	statement	lemma C.3	arxiv
Streaming-history properties	statement	lemma B.2	arxiv
Full upper frontier	statement	corollary 4.2	arxiv
Robustness to inexact history and master updates	statement	lemma B.15	arxiv
Comparator-KL monotonicity of the feasibility correction	statement	lemma B.4	arxiv
Variation partition	statement	lemma B.10	arxiv
Scale endpoints	statement	lemma B.12	arxiv
Explicit fixed-scale guarantee	statement	lemma B.11	arxiv
Low-variation interval bound	statement	lemma B.9	arxiv