confidence intervals
Confidence intervals are statistical tools that indicate the degree of uncertainty or reliability associated with a model's predictions. In AI, confidence intervals help convey the model's certainty regarding its outputs, which is essential for decision-making in uncertain environments.
- Active Measurement: Efficient Estimation at Scale
- Doubly-Robust Estimation of Counterfactual Policy Mean Embeddings
- FAIR Universe HiggsML Uncertainty Dataset and Competition
- Pareto Optimal Risk-Agnostic Distributional Bandits with Heavy-Tail Rewards
- Quantifying Uncertainty in Error Consistency: Towards Reliable Behavioral Comparison of Classifiers
- STAR-Bets: Sequential TArget-Recalculating Bets for Tighter Confidence Intervals
- Simulation-Based Inference for Adaptive Experiments
- Size-adaptive Hypothesis Testing for Fairness
- Smooth Sailing: Lipschitz-Driven Uncertainty Quantification for Spatial Associations
- Statistical Inference for Gradient Boosting Regression
- Statistical Inference under Performativity
- Statistical inference for Linear Stochastic Approximation with Markovian Noise
- Transferring Causal Effects using Proxies
- When Data Can't Meet: Estimating Correlation Across Privacy Barriers