TRAK: Attributing Model Behavior at Scale

2023 inproceedings park2023trak ○ Not yet verified

Authors

Sung Min Park, Kristian Georgiev, Andrew Ilyas, Guillaume Leclerc, Aleksander Madry

Venue

International Conference on Machine Learning (ICML)

Abstract

Introduces TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale models by leveraging random projections.

BibTeX

Local Entry

@inproceedings{park2023trak,
  title = {TRAK: Attributing Model Behavior at Scale},
  author = {Sung Min Park and Kristian Georgiev and Andrew Ilyas and Guillaume Leclerc and Aleksander Madry},
  year = {2023},
  booktitle = {International Conference on Machine Learning (ICML)},
  url = {https://arxiv.org/abs/2303.14186},
  abstract = {Introduces TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale models by leveraging random projections.}
}

From OPENALEX

@inproceedings{park2023trak,
  title = {TRAK: Attributing Model Behavior at Scale},
  author = {Sung‐Min Park and Kristian Georgiev and Andrew Ilyas and Guillaume Leclerc and Aleksander Ma̧dry},
  year = {2023},
  booktitle = {arXiv (Cornell University)},
  doi = {10.48550/arxiv.2303.14186}
}