A new optimistic online mirror descent variant uses a post-hoc penalty to allow learning rates up to Θ(T) while bounding cumulative penalty at O(log T), achieving near-optimal dynamic regret and faster adaptation on non-stationary data.
Minku, João Gama, Jerzy Stefanowski, and Michał Woźniak
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
-
Agile Online Model Selection: Resolving Adaptation Lag via Safeguarded Large Learning Rates
A new optimistic online mirror descent variant uses a post-hoc penalty to allow learning rates up to Θ(T) while bounding cumulative penalty at O(log T), achieving near-optimal dynamic regret and faster adaptation on non-stationary data.