arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2201.04234cs.LGstat.ML
收藏

利用未标记数据预测分布外性能

Leveraging Unlabeled Data to Predict Out-of-Distribution Performance

Carnegie Mellon University(卡内基梅隆大学);Google Research(谷歌研究院)

Saurabh Garg, Sivaraman Balakrishnan, Zachary C. Lipton, Behnam Neyshabur, Hanie Sedghi

摘要:

现实世界的机器学习部署的特点是源(训练)和目标(测试)分布之间的不匹配,这可能导致性能下降。在这项工作中,我们研究的方法,预测目标域的准确性,只使用标记的源数据和未标记的目标数据。我们提出了平均保留置信度(ATC),这是一种实用的方法,它可以学习模型置信度的阈值,预测精度为模型置信度超过该阈值的未标记示例的比例。ATC在几种模型架构、分布偏移类型(例如,由于合成腐败,数据集复制或新的亚群)和数据集(Wilds,ImageNet,Breeds,CIFAR和MNIST)。在我们的实验中,ATC估计目标性能$2$ - $4\times$比以前的方法更准确。我们还探讨了这个问题的理论基础,证明,在一般情况下,识别的准确性是一样困难的识别最佳预测,因此,任何方法的有效性取决于(也许未说明)的性质的转变的假设。最后,分析我们的方法在一些玩具分布,我们提供了有关它的工作时的见解。代码可在https://github.com/saurabhgarg1996/ATC_code/上获得。

英文摘要:

Real-world machine learning deployments are characterized by mismatches between the source (training) and target (test) distributions that may cause performance drops. In this work, we investigate methods for predicting the target domain accuracy using only labeled source data and unlabeled target data. We propose Average Thresholded Confidence (ATC), a practical method that learns a threshold on the model's confidence, predicting accuracy as the fraction of unlabeled examples for which model confidence exceeds that threshold. ATC outperforms previous methods across several model architectures, types of distribution shifts (e.g., due to synthetic corruptions, dataset reproduction, or novel subpopulations), and datasets (Wilds, ImageNet, Breeds, CIFAR, and MNIST). In our experiments, ATC estimates target performance $2$-$4\times$ more accurately than prior methods. We also explore the theoretical foundations of the problem, proving that, in general, identifying the accuracy is just as hard as identifying the optimal predictor and thus, the efficacy of any method rests upon (perhaps unstated) assumptions on the nature of the shift. Finally, analyzing our method on some toy distributions, we provide insights concerning when it works. Code is available at https://github.com/saurabhgarg1996/ATC_code/.

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。

↑