English

The Sample Complexity of Replicable Realizable PAC Learning

Machine Learning 2026-02-24 v1 Computational Complexity Data Structures and Algorithms

Abstract

In this paper, we consider the problem of replicable realizable PAC learning. We construct a particularly hard learning problem and show a sample complexity lower bound with a close to (logH)3/2(\log|H|)^{3/2} dependence on the size of the hypothesis class HH. Our proof uses several novel techniques and works by defining a particular Cayley graph associated with HH and analyzing a suitable random walk on this graph by examining the spectral properties of its adjacency matrix. Furthermore, we show an almost matching upper bound for the lower bound instance, meaning if a stronger lower bound exists, one would have to consider a different instance of the problem.

Keywords

Cite

@article{arxiv.2602.19552,
  title  = {The Sample Complexity of Replicable Realizable PAC Learning},
  author = {Kasper Green Larsen and Markus Engelund Mathiasen and Chirag Pabbaraju and Clement Svendsen},
  journal= {arXiv preprint arXiv:2602.19552},
  year   = {2026}
}
R2 v1 2026-07-01T10:46:56.873Z