Introduction to conformal predictors

作者：

Highlights：

• Conformal predictors output set-valued predictions with an assigned error rate (validity).

• CP require only the assumption of independent and identically distributed data.

• CP operate on top of virtually any machine learning method.

• The more accurate the underlying ML methods, the narrower the predictions (efficiency).

摘要

•Conformal predictors output set-valued predictions with an assigned error rate (validity).•CP require only the assumption of independent and identically distributed data.•CP operate on top of virtually any machine learning method.•The more accurate the underlying ML methods, the narrower the predictions (efficiency).

论文关键词：Conformal prediction,Nonparametric methods,Confidence

论文评审过程：Received 20 September 2021, Revised 18 December 2021, Accepted 22 December 2021, Available online 25 December 2021, Version of Record 4 January 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108507