Abstract

This note discusses a simple modification of cross-conformal predictioninspired by recent work on e-values. The precursor of conformal predictiondeveloped in the 1990s by Gammerman, Vapnik, and Vovk was also based one-values and is called conformal e-prediction in this note. Replacing e-valuesby p-values led to conformal prediction, which has important advantages overconformal e-prediction without obvious disadvantages. The situation withcross-conformal prediction is, however, different: whereas for cross-conformalprediction validity is only an empirical fact (and can be broken with excessiverandomization), this note draws the reader's attention to the obvious fact thatcross-conformal e-prediction enjoys a guaranteed property of validity.

 

Quick Read (beta)