Adversarial Online Learning with noise

TL;DR

This paper investigates adversarial online learning models with noisy feedback, providing tight regret bounds for scenarios with both constant and variable Bernoulli noise rates in full information and bandit settings.

Contribution

It introduces models of adversarial online learning with noisy feedback and derives tight regret bounds for these settings, extending understanding of learning under noise.

Findings

Tight regret bounds established for noisy feedback scenarios.

Analysis covers both full information and bandit feedback.

Results include constant and variable Bernoulli noise rates.

Abstract

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.