sublinear regret
A performance measure indicating the difference between the optimal solution and the algorithm's performance that grows slower than a linear rate as more data is observed.
- Fairness-Regularized Online Optimization with Switching Costs
- Greedy Algorithms for Structured Bandits: A Sharp Characterization of Asymptotic Success / Failure
- Incentivizing Truthful Language Models via Peer Elicitation Games
- No-Regret Learning Under Adversarial Resource Constraints: A Spending Plan Is All You Need!
- Online Two-Stage Submodular Maximization
- Private Online Learning against an Adaptive Adversary: Realizable and Agnostic Settings
- SOMBRL: Scalable and Optimistic Model-Based RL
- Stochastic Principal-Agent Problems: Computing and Learning Optimal History-Dependent Policies
- Taming Adversarial Constraints in CMDPs