Preprint
Aug 2026
Does 1/2-Tsallis-INF Also Work Well for Best-Arm Identification?
A Lyapunov function is constructed for the gap process between the estimated cumulative loss of the optimal arm and that of the best competing arm, and establishes a lower bound on the exponent of $\operatorname{Err}_t$ for any $\rho>0$, showing that the exponent $2$ is essentially tight.
Jingxin Zhan, Yuze Han, Zhihua Zhang
· 0 citations