Concept累积 regret / cumulative regret
regret
A core performance measure in online learning that quantifies the gap between the cumulative loss of an online learner and that of the best decision in hindsight. In the online-to-nonconvex conversion framework it corresponds directly to the quality of the stationary point in nonconvex optimization.
Timeline (last 90 days)
Oct 14
The pseudo-regret of g-Forcing satisfies R(T)=O(g(T)) and R(T)/g(T) converges almost surely to the sum of the suboptimality gaps of all suboptimal arms, but this is an asymptotic statement and does not answer the expected performance and extreme risk under finite T.
Unverified50%