Unified high-probability regret bounds for online convex optimisation with ℓq-Lipschitz losses via ℓp-regularised FTRL and cone-measure sampling from ℓr-spheres, for all p,q,r in [1,∞].
As Proposition 5.2 relies on Lipschitz concentration on theℓ1-sphere, the analogous proposition forg◦depends on the concentration of a Lipschitz function on theℓ2-sphere
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
High-probability zeroth-order online convex optimisation beyond Euclidean geometry
Unified high-probability regret bounds for online convex optimisation with ℓq-Lipschitz losses via ℓp-regularised FTRL and cone-measure sampling from ℓr-spheres, for all p,q,r in [1,∞].