Tag
该论文研究具有已知有向反馈图的对抗性多臂老虎机的一阶遗憾保证，通过固定点采样规则和两层聚合方法，将强可观察图的期望遗憾改进至接近最优的 Õ(√(αL*) + α)，弥合了最坏情况与一阶保证之间的图依赖差距。
This paper presents prior-free algorithms for improving multi-armed bandits, achieving optimal competitive ratios where scale, curvature, and horizon are free without noise, but not jointly under noise.
This paper resolves an open problem by showing that a single algorithm achieves optimal switching regret for every S against an oblivious adversary in multi-armed bandits.
This paper introduces a loss-aware feature-map pruning framework for convolutional neural networks using multi-armed bandits (UCB1 and Thompson Sampling) to selectively remove redundant feature maps, reducing computational cost while preserving accuracy.
This paper introduces a discrepancy-rounding framework for stochastic bandits with exact minimum-exposure constraints, achieving fair regret governed by the nonmandatory budget rather than horizon. It proposes algorithms with minimax and instance-dependent optimality guarantees, handles time-varying and overlapping group floors, and validates through experiments.
This paper introduces a cost-aware bandit framework for mobile crowdsensing that models worker performance as an increasing-then-converging function and handles unknown costs under budget constraints.
This paper proposes a novel structured neuron pruning framework for deep neural networks using multi-armed bandit algorithms, demonstrating effectiveness on various tasks.
This paper studies compute allocation in LLM-guided evolutionary search, identifies empirical regularities, and proposes BaSE, a multi-armed bandit algorithm that improves mean fitness and reliability across multiple models and tasks.
This paper introduces Pareto UCB1 Gossip and Simulated NSW UCB Gossip for multi-objective multi-agent multi-armed bandits, addressing both learning efficiency and fairness in stochastic environments.