Multi‐split conformal prediction via Cauchy aggregation

Conformal inference is a popular tool for constructing prediction intervals (PIs). Due to the consideration of computational burden, one most commonly used conformal methods split conformal, which generally suffers from introducing extra randomness and reducing effectiveness training models. A natural remedy use multiple splits; however, it still challenging obtain valid PIs because dependence across splits. In this paper, we propose simple yet efficient multi‐split method via adapting Cauchy aggregation, powerful combining ‐values with arbitrary correlation structures. Under two different kinds general conditions, show that our able yield asymptotically‐exact PIs. Numerical results resulting outperform existing in many settings, especially when stability condition regression modelling does not satisfy well.