Conditional validity of inductive conformal predictors

Abstract

Conformal predictors are set predictors that are automatically valid in the sense of having coverage probability equal to or exceeding a given confidence level. Inductive conformal predictors are a computationally efficient version of conformal predictors satisfying the same property of validity. However, inductive conformal predictors have been only known to control unconditional coverage probability. This paper explores various versions of conditional validity and various ways to achieve them using inductive conformal predictors and their modifications.

23 pages, 9 figures, 2 tables; to appear in the ACML 2012 Proceedings

References in corpus (2)

• Distribution Free Prediction Bands
• Efficient Nonparametric Conformal Prediction Regions

Cited by in corpus (5)

• Deep Weighted Averaging Classifiers
• Flexible distribution-free conditional predictive bands using density estimators
• Selection and Aggregation of Conformal Prediction Sets
• Sensitivity Analysis of Individual Treatment Effects: A Robust Conformal Inference Approach
• From conformal to probabilistic prediction