Learnability Beyond Uniform Convergence

Abstract

The problem of characterizing learnability is the most basic question of statistical learning theory. A fundamental result is that learnability is equivalent to uniform convergence of the empirical risk to the population risk, and that if a problem is learnable, it is learnable via empirical risk minimization. The equivalence of uniform convergence and learnability was formally established only in the supervised classification and regression setting. We show that in (even slightly) more complex prediction problems learnability does not imply uniform convergence. We discuss several alternative attempts to characterize learnability. This extended abstract summarizes results published in [5, 3].

Cite

Text

Shalev-Shwartz. "Learnability Beyond Uniform Convergence." International Conference on Algorithmic Learning Theory, 2012. doi:10.1007/978-3-642-34106-9_3

Markdown

[Shalev-Shwartz. "Learnability Beyond Uniform Convergence." International Conference on Algorithmic Learning Theory, 2012.](https://mlanthology.org/alt/2012/shalevshwartz2012alt-learnability/) doi:10.1007/978-3-642-34106-9_3

BibTeX

@inproceedings{shalevshwartz2012alt-learnability,
  title     = {{Learnability Beyond Uniform Convergence}},
  author    = {Shalev-Shwartz, Shai},
  booktitle = {International Conference on Algorithmic Learning Theory},
  year      = {2012},
  pages     = {13-16},
  doi       = {10.1007/978-3-642-34106-9_3},
  url       = {https://mlanthology.org/alt/2012/shalevshwartz2012alt-learnability/}
}