Tight First-and Second-Order Regret Bounds for Adversarial Linear Bandits

Neural Information Processing Systems 

We propose novel algorithms with first-and second-order regret bounds for adversarial linear bandits.