Combining Adversarial Guarantees and Stochastic Fast Rates in Online Learning

Abstract

We consider online learning algorithms that guarantee worst-case regret rates in adversarial environments (so they can be deployed safely and will perform robustly), yet adapt optimally to favorable stochastic environments (so they will perform well in a variety of settings of practical importance). We quantify the friendliness of stochastic environments by means of the well-known Bernstein (a.k.a. generalized Tsybakov margin) condition. For two recent algorithms (Squint for the Hedge setting and MetaGrad for online convex optimization) we show that the particular form of their data-dependent individual-sequence regret guarantees implies that they adapt automatically to the Bernstein parameters of the stochastic environment. We prove that these algorithms attain fast rates in their respective settings both in expectation and with high probability.

Cited by in corpus (11)

• The Many Faces of Exponential Weights in Online Learning
• MetaGrad: Multiple Learning Rates in Online Learning
• Impossible Tuning Made Possible: A New Expert Algorithm and Its Applications
• Fast Rates for Online Prediction with Abstention
• PAC-Bayes, MAC-Bayes and Conditional Mutual Information: Fast rate bounds that handle general VC classes
• MetaGrad: Adaptation using Multiple Learning Rates in Online Learning
• Prediction with Corrupted Expert Advice
• Fast Rates of ERM and Stochastic Approximation: Adaptive to Error Bound Conditions
• Lipschitz Adaptivity with Multiple Learning Rates in Online Learning
• Risk-Monotonicity in Statistical Learning
• Malicious Experts versus the multiplicative weights algorithm in online prediction