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.
@article{arxiv.2007.16109,
title = {Sequential Drift Detection in Deep Learning Classifiers},
author = {Samuel Ackerman and Parijat Dube and Eitan Farchi},
journal= {arXiv preprint arXiv:2007.16109},
year = {2020}
}