Adversarial Online Learning with noise
This work addresses the problem of robust online learning under noisy conditions for machine learning practitioners, but it appears incremental as it extends existing adversarial models with noise.
The paper tackles adversarial online learning with noisy feedback, considering both full-information and bandit settings with Bernoulli noise, and achieves tight regret bounds as the main result.
We present and study models of adversarial online learning where the feedback observed by the learner is noisy, and the feedback is either full information feedback or bandit feedback. Specifically, we consider binary losses xored with the noise, which is a Bernoulli random variable. We consider both a constant noise rate and a variable noise rate. Our main results are tight regret bounds for learning with noise in the adversarial online learning model.