ML Anthology

Shortest Program Interpolation Learning

Naren Sarayu Manoj, Nathan Srebro
COLT 2023 pp. 4881-4901
/colt/2023/manoj2023colt-shortest/

Abstract

We prove that the Minimum Program Length learning rule exhibits tempered overfitting. We obtain tempered agnostic finite sample learning guarantees and characterize the asymptotic behavior in the presence of random label noise.

PDF COLT Semantic Scholar

Cite

Text

Manoj and Srebro. "Shortest Program Interpolation Learning." Conference on Learning Theory, 2023.

Markdown

[Manoj and Srebro. "Shortest Program Interpolation Learning." Conference on Learning Theory, 2023.](https://mlanthology.org/colt/2023/manoj2023colt-shortest/)

BibTeX

@inproceedings{manoj2023colt-shortest,
  title     = {{Shortest Program Interpolation Learning}},
  author    = {Manoj, Naren Sarayu and Srebro, Nathan},
  booktitle = {Conference on Learning Theory},
  year      = {2023},
  pages     = {4881-4901},
  volume    = {195},
  url       = {https://mlanthology.org/colt/2023/manoj2023colt-shortest/}
}