FTRL
A classical framework in online learning in which, at each round, the decision minimizing the cumulative linear loss plus a regularization term is selected. Adjusting the regularization term flexibly controls the stability and adaptivity of the algorithm.
Timeline (last 90 days)
IAFTRL (information-acquisition follow-the-regularized-leader) is claimed by the paper to be the first linear bandit algorithm that achieves both adversarial-optimal (~O(√T) regret) and stochastic instance-dependent optimal (~O(C* log T) regret) performance within a pure FTRL framework.
IAFTRL dynamically adjusts the regularization strength, incorporating information-acquisition complexity into the design of the FTRL regularization term, thereby automatically adapting to instance difficulty.
FTRL (Follow-The-Regularized-Leader) is a classic framework in online learning, selecting at each round the decision that minimizes the cumulative linear loss plus a regularization term.
IAFTRL's analysis combines the stability of FTRL with the instance-dependent lower bound of information acquisition, providing a new tool for subsequent research.
All Facts (4)
IAFTRL (information-acquisition follow-the-regularized-leader) is claimed by the paper to be the first linear bandit algorithm that achieves both adversarial-optimal (~O(√T) regret) and stochastic instance-dependent optimal (~O(C* log T) regret) performance within a pure FTRL framework.
50%UnverifiedIAFTRL dynamically adjusts the regularization strength, incorporating information-acquisition complexity into the design of the FTRL regularization term, thereby automatically adapting to instance difficulty.
50%UnverifiedFTRL (Follow-The-Regularized-Leader) is a classic framework in online learning, selecting at each round the decision that minimizes the cumulative linear loss plus a regularization term.
50%UnverifiedIAFTRL's analysis combines the stability of FTRL with the instance-dependent lower bound of information acquisition, providing a new tool for subsequent research.
50%