Let G=(gij) be an M×N random matrix with independent standard Gaussian entries, and let
Z(G):={σ∈{−1,1}N:Gσ≥0 coordinatewise}.
This is the zero-margin binary (or Ising) perceptron. Write M=⌊αN⌋.
Define C80 to be the infimum of all α>0 such that
P(Z(G)>0)→0as N→∞.
Equivalently, C80 is the asymptotic storage-capacity / satisfiability threshold of the zero-margin Ising perceptron. Sharp-threshold results [X2021], [NS2023] show that this formulation captures the threshold location up to o(1).