English

Capacity threshold for the Ising perceptron

Probability 2025-04-23 v2 Disordered Systems and Neural Networks Mathematical Physics math.MP

Abstract

We show that the capacity of the Ising perceptron is with high probability upper bounded by the constant α0.833\alpha_\star \approx 0.833 conjectured by Krauth and M\'ezard, under the condition that an explicit two-variable function S(λ1,λ2)\mathscr{S}_*(\lambda_1,\lambda_2) is maximized at (1,0)(1,0). The earlier work of Ding and Sun proves the matching lower bound subject to a similar numerical condition, and together these results give a conditional proof of the conjecture of Krauth and M\'ezard.

Cite

@article{arxiv.2404.18902,
  title  = {Capacity threshold for the Ising perceptron},
  author = {Brice Huang},
  journal= {arXiv preprint arXiv:2404.18902},
  year   = {2025}
}

Comments

76 pages, 2 figures. This version includes rigorous interval arithmetic verification of the numerical estimates in Appendix B, carried out in the attached Python file. This proves all numerical conditions in the paper (for kappa=0) except the main Condition 1.3. We slightly reformulate Condition 3.4 to simplify the numerical verification

R2 v1 2026-06-28T16:10:08.152Z