English

Properly learning monotone functions via local reconstruction

Data Structures and Algorithms 2023-03-29 v3

Abstract

We give a 2O~(n/ϵ)2^{\tilde{O}(\sqrt{n}/\epsilon)}-time algorithm for properly learning monotone Boolean functions under the uniform distribution over {0,1}n\{0,1\}^n. Our algorithm is robust to adversarial label noise and has a running time nearly matching that of the state-of-the-art improper learning algorithm of Bshouty and Tamon (JACM '96) and an information-theoretic lower bound of Blais et al (RANDOM '15). Prior to this work, no proper learning algorithm with running time smaller than 2Ω(n)2^{\Omega(n)} was known to exist. The core of our proper learner is a \emph{local computation algorithm} for sorting binary labels on a poset. Our algorithm is built on a body of work on distributed greedy graph algorithms; specifically we rely on a recent work of Ghaffari (FOCS'22), which gives an efficient algorithm for computing maximal matchings in a graph in the LCA model of Rubinfeld et al and Alon et al (ICS'11, SODA'12). The applications of our local sorting algorithm extend beyond learning on the Boolean cube: we also give a tolerant tester for Boolean functions over general posets that distinguishes functions that are ϵ/3\epsilon/3-close to monotone from those that are ϵ\epsilon-far. Previous tolerant testers for the Boolean cube only distinguished between ϵ/Ω(n)\epsilon/\Omega(\sqrt{n})-close and ϵ\epsilon-far.

Keywords

Cite

@article{arxiv.2204.11894,
  title  = {Properly learning monotone functions via local reconstruction},
  author = {Jane Lange and Ronitt Rubinfeld and Arsen Vasilyan},
  journal= {arXiv preprint arXiv:2204.11894},
  year   = {2023}
}

Comments

FOCS 2022

R2 v1 2026-06-24T10:58:13.577Z