Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Testing Whether a Learning Procedure is Calibrated
Jon Cockayne, Matthew M. Graham, Chris J. Oates, T. J. Sullivan, Onur Teymur
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use noisy ABC instead of standard ABC | ABC on g-and-k distribution · g-and-k location inference, tolerances 1 to 10 | strong and weak calibration test (KS, p-value) | Improved |
| increase ABC tolerance to epsilon >= 9 | standard ABC · g-and-k, strong calibration test | rejection of strong calibration at 0.05 | Hurt |
| increase Student-t degrees of freedom or number of data | Laplace approximation · Student-t location model | power of calibration tests | Hurt |
| use weak calibration test instead of strong test for probabilistic ODE solvers | probabilistic ODE solvers · Lotka-Volterra | calibration rejection | No clear effect |
| select test location x_* from held-out data split instead of user-specified x=0.5 | strong calibration goodness-of-fit test · Non-stationary GP misspecified by stationary GP | average log p-value | Improved |
| increase number of simulations in the selection subset (s=10 to 150) | data-driven test function selection · Non-stationary GP | ease of identifying x_* | Improved |
Closest papers by meaning
- Combining Priors with Experience: Confidence Calibration Based on Binomial Process Modeling (AAAI 2025 · similarity 0.85)
- Calibration without labels in multiple testing (NeurIPS 2026 · similarity 0.84)
- Testing Calibration in Nearly-Linear Time (NeurIPS 2024 · similarity 0.84)
- Reassessing How to Compare and Improve the Calibration of Machine Learning Models (ICLR 2025 · similarity 0.83)
- Parametric ρ-Norm Scaling Calibration (AAAI 2025 · similarity 0.83)
- Experts Don't Cheat: Learning What You Don't Know By Predicting Pairs (ICML 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
