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 2007.16109stat.APcs.LGstat.ML
收藏

深度学习分类器中的序列漂移检测

Sequential Drift Detection in Deep Learning Classifiers

IBM Research, Israel(IBM研究院(以色列));IBM Research, USA(IBM研究院(美国))

Samuel Ackerman, Parijat Dube, Eitan Farchi

摘要:

我们利用神经网络嵌入检测数据漂移制定一个适当的顺序决策框架内的漂移检测。这使得能够控制误报警率,尽管重复应用统计测试。由于变化检测算法自然面临着避免误报和快速正确检测之间的权衡,我们引入了一个损失函数,评估算法的能力,以平衡这两个问题,我们使用它在一系列的实验。

英文摘要:

We utilize neural network embeddings to detect data drift by formulating the drift detection within an appropriate sequential decision framework. This enables control of the false alarm rate although the statistical tests are repeatedly applied. Since change detection algorithms naturally face a tradeoff between avoiding false alarms and quick correct detection, we introduce a loss function which evaluates an algorithm's ability to balance these two concerns, and we use it in a series of experiments.

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。

↑