arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.01416cs.LGstat.ML
收藏

在无标签情况下使预测集适应分布偏移

Adapting Prediction Sets to Distribution Shifts Without Labels

School of Engineering University of Guelph(工程学院 温哥华大学);Vector Institute for Artificial Intelligence(人工智能矢量研究所);Harvard University(哈佛大学)

Kevin Kasa, Zhiyu Zhang, Heng Yang, Graham W. Taylor

AI总结:

本文研究在测试期无标签且存在分布偏移时如何改进 conformal prediction (CP),提出 ECP 和 EACP 两种方法,利用基础模型自身的不确定性评估调整得分函数,实验表明其性能显著优于基线并接近全监督方法。

摘要:

近期,部署置信集预测而非点预测在机器学习中引起了广泛关注。然而,此类预测集的有效性在实践中常因分布偏移而受损,且测试期缺乏真实标签往往使该挑战进一步加剧。本文聚焦于一种名为 conformal prediction (CP) 的标准集值预测框架,研究如何仅利用来自偏移测试域的无标签数据来改善其实际性能。这一目标通过两种名为 ECP 和 EACP 的新方法实现,其核心思想是根据 CP 中基础模型自身的不确定性评估来调整得分函数。通过在多个大规模数据集和神经网络架构上的广泛实验,我们表明所提方法相较现有基线提供了一致的改进,并且几乎匹配了全监督方法的性能。

英文摘要:

Recently there has been a surge of interest to deploy confidence set predictions rather than point predictions in machine learning. Unfortunately, the effectiveness of such prediction sets is frequently impaired by distribution shifts in practice, and the challenge is often compounded by the lack of ground truth labels at test time. Focusing on a standard set-valued prediction framework called conformal prediction (CP), this paper studies how to improve its practical performance using only unlabeled data from the shifted test domain. This is achieved by two new methods called ECP and EACP, whose main idea is to adjust the score function in CP according to its base model's own uncertainty evaluation. Through extensive experiments on a number of large-scale datasets and neural network architectures, we show that our methods provide consistent improvement over existing baselines and nearly match the performance of fully supervised methods.

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。

↑