Corpus ID: 220364464

Linear Bandits with Limited Adaptivity and Learning Distributional Optimal Design

  title={Linear Bandits with Limited Adaptivity and Learning Distributional Optimal Design},
  author={Yufei Ruan and Jiaqi Yang and Y. Zhou},
  • Yufei Ruan, Jiaqi Yang, Y. Zhou
  • Published 2020
  • Computer Science, Mathematics
  • ArXiv
  • Motivated by practical needs such as large-scale learning, we study the impact of adaptivity constraints to linear contextual bandits, a central problem in online active learning. We consider two popular limited adaptivity models in literature: batch learning and rare policy switches. We show that, when the context vectors are adversarially chosen in $d$-dimensional linear contextual bandits, the learner needs $O(d \log d \log T)$ policy switches to achieve the minimax-optimal regret, and this… CONTINUE READING
    4 Citations
    Provable Benefits of Representation Learning in Linear Bandits
    • 1
    • PDF
    Double Explore-then-Commit: Asymptotic Optimality and Beyond
    • 2
    • PDF
    Batched Neural Bandits
    • Highly Influenced
    • PDF


    Sequential Batch Learning in Finite-Action Linear Contextual Bandits
    • 10
    • PDF
    Bandits with switching costs: T2/3 regret
    • 47
    • PDF
    Nearly Minimax-Optimal Regret for Linearly Parameterized Bandits
    • 15
    • Highly Influential
    • PDF
    Provably Efficient Q-Learning with Low Switching Cost
    • 15
    • PDF
    Online Learning with Switching Costs and Other Adaptive Adversaries
    • 57
    • PDF
    Stochastic Linear Optimization under Bandit Feedback
    • 515
    • PDF
    Linearly Parameterized Bandits
    • 324
    • PDF
    Contextual Bandits with Linear Payoff Functions
    • 510
    • Highly Influential
    • PDF
    Minimax Bounds on Stochastic Batched Convex Optimization
    • 17
    • PDF