Conceptminimax regret
Minimax Regret
A standard metric for measuring bandit algorithm performance, defined as an upper bound on the difference between the algorithm's cumulative loss and the cumulative loss of the best fixed action, taken over the worst case of the environment.
Timeline (last 90 days)
Nov 4
For weakly observable graph bandits, the known minimax regret lower bound is Θ(T^{2/3}), where T is the total number of rounds.
Unverified50%