A second order regret bound for NormalHedge

2026-02-08

AI Summary

This work addresses the challenge of optimizing higher-order regret bounds in expert prediction by proposing an improved NormalHedge algorithm. By leveraging insights from continuous-time stochastic differential equations to guide algorithm design and combining them with self-concordant analysis techniques in discrete time, the paper establishes, for the first time, a second-order ε-quantile regret bound for NormalHedge. Under the condition that the cumulative loss variance satisfies \(V_T > \log N\), the resulting regret bound is \(O(\sqrt{V_T \log(V_T/\varepsilon)})\), which significantly improves upon conventional first-order bounds—particularly in high-variance regimes where the advantage becomes especially pronounced.

Technology Category

• Machine Learning: Online Learning & Bandits
• Reasoning under Uncertainty: Stochastic Optimization
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

We consider the problem of prediction with expert advice for ``easy''sequences. We show that a variant of NormalHedge enjoys a second-order $\epsilon$-quantile regret bound of $O\big(\sqrt{V_T \log(V_T/\epsilon)}\big) $ when $V_T>\log N$, where $V_T$ is the cumulative second moment of instantaneous per-expert regret averaged with respect to a natural distribution determined by the algorithm. The algorithm is motivated by a continuous time limit using Stochastic Differential Equations. The discrete time analysis uses self-concordance techniques.

Problem

• prediction with expert advice
• regret bound
• second-order regret
• easy sequences
• NormalHedge

Innovation

• second-order regret
• NormalHedge
• stochastic differential equations
• self-concordance
• prediction with expert advice