Root-selected quantitative AC0 parity lower bound #
This module selects a canonical scale from Mathlib's verified natural-number
nthRoot. Let
q = Nat.nthRoot (d - 1) n / 20
and use t = q - 1. Once 40^(d-1) <= n, one has q >= 2 and
20 * (t + 1) <= Nat.nthRoot (d - 1) n, so the scale theorem applies. Every
depth-d parity circuit, even with arbitrary internal NOT gates, then satisfies
2^q <= 20 * (q - 1) * S.
This is a conventional explicit form of the
exp(Omega(n^(1/(d-1)))) lower bound, with conservative constants chosen for
the exact first-moment proof. Root selection is symbolic and uniform; it does
not enumerate circuits or run bounded experiments.
The integer root, divided by the scale constant.
Equations
- Algebraic.AC0.ParityParameters.rootQuotient n depth = (depth - 1).nthRoot n / 20
Instances For
Target tree depth obtained by reserving one unit below the root quotient.
Equations
Instances For
Root-selected product-form lower bound for depth-d parity circuits,
allowing arbitrary internal NOT gates at zero cost.
Compatibility wrapper for the checked input-negation presentation.
Root-selected lower bound with the AND/OR cost isolated by natural-number floor division, allowing arbitrary internal NOT gates at zero cost.
Compatibility wrapper for the checked input-negation presentation.