The limits of distribution-free conditional predictive inference
From MaRDI portal
Abstract: We consider the problem of distribution-free predictive inference, with the goal of producing predictive coverage guarantees that hold conditionally rather than marginally. Existing methods such as conformal prediction offer marginal coverage guarantees, where predictive coverage holds on average over all possible test points, but this is not sufficient for many practical applications where we would like to know that our predictions are valid for a given individual, not merely on average over a population. On the other hand, exact conditional inference guarantees are known to be impossible without imposing assumptions on the underlying distribution. In this work we aim to explore the space in between these two, and examine what types of relaxations of the conditional coverage property would alleviate some of the practical concerns with marginal coverage guarantees while still being possible to achieve in a distribution-free setting.
Recommendations
- Distribution-free predictive inference for regression
- Training-conditional coverage for distribution-free predictive inference
- A refinement to approximate conditional inference
- Model-free model-fitting and predictive distributions
- Conditional inferential models: combining information for prior-free probabilistic inference
- Prediction by quantization of a conditional distribution
- Conditional predictive inference for stable algorithms
- Limit theory and inference about conditional distributions
- Distribution-free conditional median inference
- Nonparametric Estimation of Conditional Distributions
Cited in
(39)- Distribution-free predictive inference for regression
- A minimax framework for quantifying risk-fairness trade-off in regression
- Robust prediction interval estimation for Gaussian processes by cross-validation method
- Conformal prediction: a unified review of theory and new challenges
- Local permutation tests for conditional independence
- Is distribution-free inference possible for binary regression?
- Distribution-free conditional median inference
- Nested conformal prediction sets for classification with applications to probation data
- Valid model-free prediction of future insurance claims
- Distribution-free, Risk-controlling Prediction Sets
- Conformal Prediction: A Gentle Introduction
- Conditional predictive inference for stable algorithms
- Testing for outliers with conformal p-values
- Selective Prediction-Set Models with Coverage Rate Guarantees
- Distribution-Free Prediction Sets for Two-Layer Hierarchical Models
- Conformal Sensitivity Analysis for Individual Treatment Effects
- Training-conditional coverage for distribution-free predictive inference
- Selective review of biased sampling problems with applications in modern statistics
- Regression trees for fast and adaptive prediction intervals
- Robust Validation: Confident Predictions Even When Distributions Shift
- Conditionality principle under unconstrained randomness
- Fair conformal prediction for incomplete covariate data
- Conditional validity of heteroskedastic conformal regression
- Predictive inference in multi-environment scenarios
- Conformal Prediction for Network-Assisted Regression
- Confidence on the focal: conformal prediction with selection-conditional coverage
- Conformal prediction with conditional guarantees
- Conformal prediction with local weights: randomization enables robust guarantees
- Conformal prediction for multivariate responses with Euclidean likelihood
- Conformal inference for random objects
- Belief in dependence: leveraging atomic linearity in data bits for rethinking generalized linear models
- Learn then test: calibrating predictive algorithms to achieve risk control
- Probabilistic scoring lists for interpretable machine learning
- Distribution-Free Prediction Intervals Under Covariate Shift, With an Application to Causal Inference
- Selection and Aggregation of Conformal Prediction Sets
- Conformal link prediction for false discovery rate control
- Online Auction Design Using Distribution-Free Uncertainty Quantification with Applications to E-Commerce
- The limits of assumption-free tests for algorithm performance
- Conformal prediction for dose-response models with continuous treatments
This page was built for publication: The limits of distribution-free conditional predictive inference
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5155952)
