1 CSAIL, MIT | 2 DataHeroes | 3 Rice University

Alaa Maalouf, Murad Tukan, Vladimir Braverman, and Daniela Rus

• Paper Link: https://arxiv.org/abs/2305.11980

• Explainer video: https://icml.cc/virtual/2023/poster/24432

A coreset is a small weighted subset that approximates the loss function on the whole data, prevalent in machine learning for its advantages. However, current construction methods are problem-dependent and may be challenging for new researchers.

No worries, we got you. We propose AutoCoreset: an automatic practical framework for constructing coresets requiring only input data and the cost function (without any other user computation or calculation), making it user-friendly and applicable to various problems. See our open-source code which supports future research and simplifies coreset usage.

To use AutoCoreset on your data and desired ML model:

(1) Modify Line - 134 to be a list containing the path to your Dataset.

(2) Modify Line - 138 to be the name of your ML model (currently we support 'k_means', 'logistic_regression', 'linear_regression', 'svm').

(3) You can certainly change the ML model as you wish - as long as you provide the "fit" and "score" functions.

(4) Run: python main.py

If you find this work helpful please cite us:

@article{maalouf2023autocoreset,
      title={AutoCoreset: An Automatic Practical Coreset Construction Framework},
      author={Maalouf, Alaa and Tukan, Murad and Braverman, Vladimir and Rus, Daniela},
      journal={arXiv preprint arXiv:2305.11980},
      year={2023}
}