arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2411.01596stat.MLcs.LG
收藏

策略性共形预测

Strategic Conformal Prediction

Fundação Getulio Vargas(热图利奥·瓦加斯基金会);School of Applied Mathematics(应用数学学院)

Daniel Csillag, Claudio José Struchiner, Guilherme Tegoni Goedert

AI总结:

提出策略性共形预测框架,在智能体策略性改变环境下实现稳健的不确定性量化,具备多项理论保证并实验验证其有效性。

摘要:

当机器学习模型被部署时,其预测可能会改变其环境,因为信息更充分的智能体会为了自身利益而采取策略性行动。考虑到这些改变,现有的不确定性量化方法会失效。在这项工作中,我们提出了一个新框架——策略性共形预测(Strategic Conformal Prediction),它能够在这种情境下进行稳健的不确定性量化。策略性共形预测得到了一系列理论保证的支持,这些保证涵盖边际覆盖、训练条件覆盖、紧致性以及对错误设定的稳健性,并且以无分布方式成立。实验分析进一步验证了我们的方法,展示了其在面对任意策略性改变时的显著有效性,而其他方法则会失效。

英文摘要:

When a machine learning model is deployed, its predictions can alter its environment, as better informed agents strategize to suit their own interests. With such alterations in mind, existing approaches to uncertainty quantification break. In this work we propose a new framework, Strategic Conformal Prediction, which is capable of robust uncertainty quantification in such a setting. Strategic Conformal Prediction is backed by a series of theoretical guarantees spanning marginal coverage, training-conditional coverage, tightness and robustness to misspecification that hold in a distribution-free manner. Experimental analysis further validates our method, showing its remarkable effectiveness in face of arbitrary strategic alterations, whereas other methods break.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑