regret bounds
Regret bounds provide theoretical guarantees on the performance difference between an algorithm's decisions and those of the best possible decision in hindsight, useful in online learning scenarios.
- An Ellipsoid Algorithm for Online Convex Optimization
- Comparator-Adaptive $\Phi$-Regret: Improved Bounds, Simpler Algorithms, and Applications to Games
- Generator-Mediated Bandits: Thompson Sampling for GenAI-Powered Adaptive Interventions
- Improved Regret and Contextual Linear Extension for Pandora's Box and Prophet Inequality
- Instance-Dependent Regret Bounds for Nonstochastic Linear Partial Monitoring
- Learning Across the Gap: Hybrid Multi-armed Bandits with Heterogeneous Offline and Online Data
- Prediction with expert advice under additive noise
- Prompt Tuning Decision Transformers with Structured and Scalable Bandits
- Replicable Online Learning
- Robust Satisficing Gaussian Process Bandits Under Adversarial Attacks
- Taming Adversarial Constraints in CMDPs
- Theoretical Guarantees for the Retention of Strict Nash Equilibria by Coevolutionary Algorithms
- Thompson Sampling in Function Spaces via Neural Operators