一句话：提出 HybridRepair，以低标注代价修复因数据分布偏移而性能下降的已部署深度学习模型。
深度学习模型在训练数据与运行环境中的现场数据分布不一致时，部署后往往无法达到预期性能，因此模型修复至关重要，尤其是当模型被部署到规模不断增大且分布发生偏移的任务上时。针对这一需求，论文提出 HybridRepair，目标是以标注高效（annotation-efficient）的方式完成深度学习模型的修复，即在修复过程中尽量减少对人工标注的依赖。摘要指出该问题源于训练分布与现场分布之间的失配，并将修复视为部署阶段的关键环节；由于给出的摘要篇幅有限，文中未说明具体的混合修复策略、所用基准任务以及实验结果的定量指标。
A well-trained deep learning (DL) model often cannot achieve expected performance after deployment due to the mismatch between the distributions of the training data and the field data in the operational environment. Therefore, repairing DL models is critical, especially when deployed on increasingly larger tasks with shifted distributions.
DOI 原文 ·
@inproceedings{paperbot7425,
title = {HybridRepair: towards annotation-efficient repair for deep learning models},
author = {Yu Li and Muxi Chen and Qiang Xu},
booktitle = {ISSTA},
year = {2022},
doi = {10.1145/3533767.3534408}
}