Localized conformal model selection

TL;DR

This paper introduces a localized conformal model selection method that adaptively constructs prediction intervals with guaranteed coverage, effectively handling heterogeneity and reducing interval length.

Contribution

It develops a novel ensemble framework that combines local adaptivity with conformal inference, ensuring valid distribution-free prediction intervals in complex settings.

Findings

Significant reduction in interval length compared to fixed models

Maintains exact finite-sample coverage in diverse scenarios

Effective in heterogeneous and low-noise environments

Abstract

We propose a localized conformal model selection framework that integrates local adaptivity with post-selection validity for distribution-free prediction. By performing model selection symmetrically across calibration points using upper and lower surrogate intervals, we construct a data-dependent safe index set that contains the oracle model and preserves exchangeability. The resulting ensemble procedure retains exact finite-sample marginal coverage while adapting to spatial heterogeneity and model complexity. Simulations demonstrate substantial reductions in interval length compared to the best fixed model, especially in heterogeneous and low-noise settings.