2018 (modified: 11 Nov 2022)ICML 2018Readers: Everyone
Abstract:We present an extensive study of a key problem in online learning where the learner can opt to abstain from making a prediction, at a certain cost. In the adversarial setting, we show how existing ...