Adaptive Smooth Non-Stationary Bandits

Joe Suk

We study a $K$-armed non-stationary bandit model where rewards change smoothly, as captured by Hölder class assumptions on rewards as functions of time. Such smooth changes are par…