k-lazyGD achieves optimal dynamic regret O(sqrt((P_T+1)T)) in SOCO for laziness k up to Theta(sqrt(T/P_T)).
In thek-lazy case, the same principle applies phase by phase, with the bound expressed relative to the average within each lazy block
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
-
Partially Lazy Gradient Descent for Smoothed Online Learning
k-lazyGD achieves optimal dynamic regret O(sqrt((P_T+1)T)) in SOCO for laziness k up to Theta(sqrt(T/P_T)).