Sequential Harmful Shift Detection Without Labels

Abstract

We introduce a novel approach for detecting distribution shifts that negatively impact the performance of machine learning models in continuous production environments, which requires no access to ground truth data labels. It builds upon the work of Podkopaev and Ramdas [2022], who address scenarios where labels are available for tracking model errors over time. Our solution extends this framework to work in the absence of labels, by employing a proxy for the true error. This proxy is derived using the predictions of a trained error estimator. Experiments show that our method has high power and false alarm control under various distribution shifts, including covariate and label shifts and natural shifts over geography and time.

PDF NeurIPS OpenReview Semantic Scholar

Cite

Text

Amoukou et al. "Sequential Harmful Shift Detection Without Labels." Neural Information Processing Systems, 2024. doi:10.52202/079017-4107

Markdown

[Amoukou et al. "Sequential Harmful Shift Detection Without Labels." Neural Information Processing Systems, 2024.](https://mlanthology.org/neurips/2024/amoukou2024neurips-sequential/) doi:10.52202/079017-4107

BibTeX

@inproceedings{amoukou2024neurips-sequential,
  title     = {{Sequential Harmful Shift Detection Without Labels}},
  author    = {Amoukou, Salim I. and Bewley, Tom and Mishra, Saumitra and Lecue, Freddy and Magazzeni, Daniele and Veloso, Manuela},
  booktitle = {Neural Information Processing Systems},
  year      = {2024},
  doi       = {10.52202/079017-4107},
  url       = {https://mlanthology.org/neurips/2024/amoukou2024neurips-sequential/}
}