finite-sample guarantees
Theoretical assurances about the performance and behavior of learning algorithms based on a limited number of samples, providing bounds on accuracy and generalization capabilities in practical settings.
- Assessing the quality of denoising diffusion models in Wasserstein distance: noisy score and optimal bounds
- Causality Meets Locality: Provably Generalizable and Scalable Policy Learning for Networked Systems
- Conformal Arbitrage: Risk-Controlled Balancing of Competing Objectives in Language Models
- Conformal Prediction for Causal Effects of Continuous Treatments
- Non-exchangeable Conformal Prediction with Optimal Transport: Tackling Distribution Shift with Unlabeled Data
- Online robust locally differentially private learning for nonparametric regression
- Path-specific effects for pulse-oximetry guided decisions in critical care
- Performative Risk Control: Calibrating Models for Reliable Deployment under Performativity
- Spectral Learning for Infinite-Horizon Average-Reward POMDPs