arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2207.09239cs.LGstat.ML
收藏

迁移学习中的分布外泛化分析

Assaying Out-Of-Distribution Generalization in Transfer Learning

AWS Tübingen(亚马逊云科技图宾根分部);Technical University of Denmark(丹麦技术大学)
Florian Wenzel, Andrea Dittadi, Peter Vincent Gehler, Carl-Johann Simon-Gabriel, Max Horn, Dominik Zietlow, David Kernert, Chris Russell, Thomas Brox, Bernt Sch… 展开作者

Florian Wenzel, Andrea Dittadi, Peter Vincent Gehler, Carl-Johann Simon-Gabriel, Max Horn, Dominik Zietlow, David Kernert, Chris Russell, Thomas Brox, Bernt Schiele, Bernhard Schölkopf, Francesco Locatello

摘要:

由于分布外泛化通常是不适定问题,因此各种代理目标(例如,校准、对抗鲁棒性、算法破坏、移位不变性)在不同的研究程序中进行了研究,得到了不同的建议。虽然这些方法有着相同的理想目标,但从未在相同的实验条件下对真实数据进行过测试。在本文中,我们对以前的工作进行了统一的看法,强调了我们经验性地解决的消息差异,并就如何衡量模型的鲁棒性以及如何改进它提供了建议。为此,我们收集了172个公开可用的数据集对,用于训练和准确性,校准误差,对抗攻击,环境不变性和合成腐败的分布外评估。我们微调了超过31k个网络,来自9种不同的架构,采用多镜头和Few-Shot设置。我们的研究结果证实,在和出分布的准确性往往会增加联合,但表明,他们的关系在很大程度上是依赖于网络,一般来说,更细致入微,更复杂的比以前的,较小规模的研究假设。

英文摘要:

Since out-of-distribution generalization is a generally ill-posed problem, various proxy targets (e.g., calibration, adversarial robustness, algorithmic corruptions, invariance across shifts) were studied across different research programs resulting in different recommendations. While sharing the same aspirational goal, these approaches have never been tested under the same experimental conditions on real data. In this paper, we take a unified view of previous work, highlighting message discrepancies that we address empirically, and providing recommendations on how to measure the robustness of a model and how to improve it. To this end, we collect 172 publicly available dataset pairs for training and out-of-distribution evaluation of accuracy, calibration error, adversarial attacks, environment invariance, and synthetic corruptions. We fine-tune over 31k networks, from nine different architectures in the many- and few-shot setting. Our findings confirm that in- and out-of-distribution accuracies tend to increase jointly, but show that their relation is largely dataset-dependent, and in general more nuanced and more complex than posited by previous, smaller scale studies.

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。

↑