PulseAugur
EN
LIVE 04:19:18

New bandit algorithms research tackles heavy tails and non-stationarity · 4 sources tracked

By PulseAugur Editorial · [6 sources] ·

Three new research papers explore advancements in bandit algorithms. One paper analyzes the regret of Thompson sampling in linear-Gaussian bandits, showing a decoupling of prior-dependent and minimax regret terms. Another paper introduces a unified misspecification-reduction approach for non-stationary linear bandits with round-specific feasible decision sets, achieving optimal dynamic-regret dependence. The third paper addresses batched multi-armed bandit problems with heavy-tailed rewards, revealing that heavier tails can surprisingly require fewer batches for near-optimal regret in certain settings. AI

IMPACT These papers advance theoretical understanding and algorithmic approaches for decision-making under uncertainty, potentially improving applications in areas like online advertising and clinical trials.

RANK_REASON Cluster consists of multiple academic papers published on arXiv concerning bandit algorithms.

Read on arXiv cs.LG →

AI-generated summary · Google Gemini · from 6 sources. How we write summaries →

New bandit algorithms research tackles heavy tails and non-stationarity · 4 sources tracked

COVERAGE [6]

  1. arXiv cs.LG TIER_1 English(EN) · Tianshuo Zheng, Ting Wu, Zhi-Hua Zhou, Keqin Liu ·

    Nonlinear Bandit

    arXiv:2607.07304v1 Announce Type: new Abstract: In this paper we first study the problem of generalized linear bandit (GLB) under heavy-tailed noise. The characteristics of heavy-tailed distributions are widely observed in real-world applications such as personalized recommendati…

  2. arXiv cs.LG TIER_1 English(EN) · Keqin Liu ·

    Nonlinear Bandit

    In this paper we first study the problem of generalized linear bandit (GLB) under heavy-tailed noise. The characteristics of heavy-tailed distributions are widely observed in real-world applications such as personalized recommendation, financial markets, and medical treatments. B…

  3. arXiv cs.LG TIER_1 English(EN) · Yifan Zhu, John C. Duchi, Benjamin Van Roy ·

    Prior Diffusiveness and Regret in the Linear-Gaussian Bandit

    arXiv:2601.02022v2 Announce Type: replace Abstract: We prove that Thompson sampling exhibits $\tilde{O}(\sigma d \sqrt{T} + d r \sqrt{\mathrm{Tr}(\Sigma_0)})$ Bayesian regret in the linear-Gaussian bandit with a $\mathcal{N}(\mu_0, \Sigma_0)$ prior distribution on the coefficient…

  4. arXiv stat.ML TIER_1 English(EN) · Zihao Hu, Yuan Yao, Jiheng Zhang, Zhengyuan Zhou ·

    Dynamic Regret for Non-Stationary Linear Bandits via Misspecification Reductions

    arXiv:2607.02891v1 Announce Type: cross Abstract: Many online decision-making problems involve both round-specific feasible actions and drifting reward models: eligible ad impressions, feasible prices, and available treatments can change over time, while user preferences, demand …

  5. arXiv stat.ML TIER_1 English(EN) · Yunwen Guo, Yunlun Shu, Gongyi Zhuo, Tianyu Wang ·

    Batched Bandits with Heavy-Tailed Rewards

    arXiv:2510.03798v3 Announce Type: replace-cross Abstract: The batched multi-armed bandit (MAB) problem, where rewards are collected in batches, is pivotal in applications like clinical trials. While prior work assumes light-tailed reward distributions, real-world scenarios often …

  6. arXiv stat.ML TIER_1 English(EN) · Zhengyuan Zhou ·

    Dynamic Regret for Non-Stationary Linear Bandits via Misspecification Reductions

    Many online decision-making problems involve both round-specific feasible actions and drifting reward models: eligible ad impressions, feasible prices, and available treatments can change over time, while user preferences, demand curves, and patient responses may evolve. Motivate…

RELATED ENTITIES

RELATED TOPICS