Unconstrained Online Linear Learning in Hilbert Spaces: Minimax Algorithms and Normal Approximations

Abstract

We study algorithms for online linear optimization in Hilbert spaces, focusing on the case where the player is unconstrained. We develop a novel characterization of a large class of minimax algorithms, recovering, and even improving, several previous results as immediate corollaries. Moreover, using our tools, we develop an algorithm that provides a regret bound of $\mathcal{O}\left(U\sqrt{T\log (U\sqrt{T}{\log }^{2}T+1)}\right)$, where $U$ is the $L_2$ norm of an arbitrary comparator and both $T$ and $U$ are unknown to the player. This bound is optimal up to $\sqrt{\log \log T}$ terms. When $T$ is known, we derive an algorithm with an optimal regret bound (up to constant factors). For both the known and unknown $T$ case, a Normal approximation to the conditional value of the game proves to be the key analysis tool.

Proceedings of the 27th Annual Conference on Learning Theory (COLT 2014)

References in corpus (2)

• No-Regret Algorithms for Unconstrained Online Convex Optimization
• A Generalized Online Mirror Descent with Applications to Classification and Regression

Cited by in corpus (19)

• Online Learning: A Modern Introduction Using Convex Optimization
• Simultaneous Model Selection and Optimization through Parameter-free Stochastic Learning
• A Survey of Algorithms and Analysis for Adaptive Online Learning
• Parameter-Free Online Convex Optimization with Sub-Exponential Noise
• A Drifting-Games Analysis for Online Learning and Applications to Boosting
• Lipschitz and Comparator-Norm Adaptivity in Online Learning
• Adaptive Online Learning
• Achieving All with No Parameters: Adaptive NormalHedge
• A Generalized Online Mirror Descent with Applications to Classification and Regression
• Training Deep Networks without Learning Rates Through Coin Betting
• Online Learning with Imperfect Hints
• Combining Online Learning Guarantees
• Matrix-Free Preconditioning in Online Learning
• Parameter-free Stochastic Optimization of Variationally Coherent Functions
• Online learning with kernel losses
• Minimizing Regret of Bandit Online Optimization in Unconstrained Action Spaces
• Comparator-adaptive Convex Bandits
• Parameter-free Gradient Temporal Difference Learning
• Adversarial Tracking Control via Strongly Adaptive Online Learning with Memory