Parameter-free algorithms for unconstrained online learning achieve regret bounds of order O(||u|| sqrt(V_T(u)) + L||u||^2 + G^4) for L-smooth convex losses without prior knowledge of ||u||, L or G, with extensions to dynamic regret and the SEA model.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Gradient-Variation Regret Bounds for Unconstrained Online Learning
Parameter-free algorithms for unconstrained online learning achieve regret bounds of order O(||u|| sqrt(V_T(u)) + L||u||^2 + G^4) for L-smooth convex losses without prior knowledge of ||u||, L or G, with extensions to dynamic regret and the SEA model.