Inductive Venn-Abers and related regressors

2026-05-07

AI Summary

This work extends the Venn-Abers predictor—previously limited to binary classification and bounded regression—to the unbounded regression setting for the first time. By incorporating a conformal prediction framework, the authors develop an inductive Venn-Abers regressor and derive from it an efficient point predictor. The proposed method preserves theoretical validity while substantially improving predictive efficiency. Empirical evaluations on both simulated and real-world datasets demonstrate that, with sufficiently large training sets, the approach consistently outperforms standard regression methods. This advancement establishes a novel paradigm for probabilistic prediction in unbounded regression tasks, offering a compelling balance between reliability and computational efficiency.

Abstract

Venn-Abers predictors are probabilistic predictors that enjoy appealing properties of validity, but their major limitation is that they are applicable only to the case of binary classification, with a recent extension to bounded regression. We generalize them to the case of unbounded regression, which requires adding an element of conformal prediction. In our simulation and empirical studies we investigate the predictive efficiency of point regressors derived from Venn-Abers regressors and argue that they somewhat improve the predictive efficiency of standard regressors for larger training sets.

Problem

• Venn-Abers predictors
• unbounded regression
• conformal prediction
• probabilistic prediction
• predictive efficiency

Innovation

• Venn-Abers predictors
• conformal prediction
• unbounded regression
• probabilistic prediction
• predictive efficiency