regret bound
A regret bound quantifies the maximum expected loss an online learning algorithm incurs compared to the best fixed strategy in hindsight, providing benchmarks for evaluating learning performance.
- A Near-optimal, Scalable and Parallelizable Framework for Stochastic Bandits Robust to Adversarial Corruptions and Beyond
- Agnostic Continuous-Time Online Learning
- An Improved Algorithm for Adversarial Linear Contextual Bandits via Reduction
- Contextual Thompson Sampling via Generation of Missing Data
- Efficient and Near-Optimal Algorithm for Contextual Dueling Bandits with Offline Regression Oracles
- Follow-the-Perturbed-Leader Nearly Achieves Best-of-Both-Worlds for the m-Set Semi-Bandit Problems
- From Contextual Combinatorial Semi-Bandits to Bandit List Classification: Improved Sample Complexity with Sparse Rewards
- No-Regret Thompson Sampling for Finite-Horizon Markov Decision Processes with Gaussian Processes
- Online Inverse Linear Optimization: Efficient Logarithmic-Regret Algorithm, Robustness to Suboptimality, and Lower Bound
- Spectral Learning for Infinite-Horizon Average-Reward POMDPs
- Tractable Multinomial Logit Contextual Bandits with Non-Linear Utilities
- Uniform Wrappers: Bridging Concave to Quadratizable Functions in Online Optimization
- Universal Sequence Preconditioning
- When Lower-Order Terms Dominate: Adaptive Expert Algorithms for Heavy-Tailed Losses