Adversarial Online Learning with Temporal Feedback Graphs

Khashayar Gatmiry, Jon Schneider

We study a variant of prediction with expert advice where the learner's action at round $t$ is only allowed to depend on losses on a specific subset of the rounds (where the struct…