共型预测方法的发展和应用

Development and application of conformal prediction

共型预测通过将预测点扩展成预测集的方式来严谨量化预测的不确定性,它以灵活的结构和严格的有限样本理论保证而著称,可以简单而方便地嵌入几乎任何预测模型中,并随着机器学习的飞速发展而在近年来获得了越来越多的关注.本文综述共型预测相关的发展历程,在回顾共型预测的基础算法和理论同时,也对共型预测无处不在的应用场景进行了介绍.

Conformal prediction has gained increasing attention in recent years with the rapid development of machine learning.Known for its flexible structure and strict finite-sample theoretical guarantees,conformal prediction can be quickly and conveniently embedded into almost any prediction model.It performs rigorous uncertainty quantification by expanding prediction points into prediction sets.In this paper,we summarize the development history related to conformal prediction and review the basic algorithms and generalizations of conformal prediction along with the ubiquitous application scenarios of conformal prediction.