跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2301.04378 · v4

Loss-Controlling Calibration with Asymptotic Guarantees

首次提交 2023-01-11 17:44
版本更新 2026-11-18 18:02
雷达观测 2026-11-19 15:04
arXiv ↗PDF ↗

摘要

We propose a learning framework for calibrating predictive models to perform loss-controlling prediction, which extends our proposed conformal loss-controlling prediction framework to more general settings. In contrast to existing methods, the predictors built by the proposed approach are not restricted to set-valued predictors, and the loss function can be any measurable function without requiring monotonicity assumptions. The approach is a natural extension of conformal loss-controlling prediction, since it reduces to the latter when the predictors are set-valued with the nesting property and the losses are monotone. The loss-controlling guarantee of the proposed framework is established in the asymptotic setting and is empirically verified in selective regression and high-impact weather forecasting, demonstrating its effectiveness for the domains requiring general loss-controlling prediction.

作者备注:Preprint

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v4,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v4 · 2026-11-18 18:02 · 最新观测版本

Loss-Controlling Calibration with Asymptotic Guarantees

We propose a learning framework for calibrating predictive models to perform loss-controlling prediction, which extends our proposed conformal loss-controlling prediction framework to more general settings. In contrast to existing methods, the predictors built by the proposed approach are not restricted to set-valued predictors, and the loss function can be any measurable function without requiring monotonicity assumptions. The approach is a natural extension of conformal loss-controlling prediction, since it reduces to the latter when the predictors are set-valued with the nesting property and the losses are monotone. The loss-controlling guarantee of the proposed framework is established in the asymptotic setting and is empirically verified in selective regression and high-impact weather forecasting, demonstrating its effectiveness for the domains requiring general loss-controlling prediction.

查看此版本解读 →查看此版本原文 ↗