Native Resource Bounds for the Bounded-Incidence Branch Evaluator (Proposition 5.3) #
This file establishes the native multi-tape Turing machine resource certificate, global polynomial
workspace bound across all raw BranchEvalInput tuples, and subexponential
$O^*(2^{O(\mathrm{interpolationScale}(n'))})$ runtime bound for the Proposition 5.3
bounded-incidence, bounded-width $\mathbb{F}_2$ polynomial-system branch evaluator
evalBoundedIncidenceBranch of OpenAI, The Exponential Time Hypothesis is False
(§§3.4–3.5, §4.7, §§5.3–5.5, Proposition 5.3, and Appendix B.2–B.3, pp. 24–26, 34–35, 37–39,
48–50).
Main definitions #
Cslib.Complexity.SubexponentialSAT.SparsePolyCounter.boundedIncidenceBranchCertificate: uniformNativeCertificateforevalBoundedIncidenceBranchonencBranchEvalInput.
Main results #
SparsePolyCounter.auxDim_lt_self_of_two_le:ReplacementParams.auxDim n' < n'for2 ≤ n'.SparsePolyCounter.polyBounded_boundedIncidenceBranchCertificate_space: globalPolyBoundedworkspace bound across all rawBranchEvalInputtuples.SparsePolyCounter.exists_boundedIncidenceBranchCertificate_time_bound: subexponential $O^*(2^{O(\mathrm{interpolationScale}(n'))})$ branch runtime bound onprepareBranchEvalInput (n', sys')under polylogarithmic factor incidence and scope width (Proposition 5.3, pp. 37–39).SparsePolyCounter.branchPolylogTimeBound_boundedIncidenceBranchCertificate: dischargesCslib.Complexity.BranchedSparseCNFCounter.BranchPolylogTimeBoundforboundedIncidenceBranchCertificate.
References #
- OpenAI, The Exponential Time Hypothesis is False, §§3.4–3.5, §4.7, §§5.3–5.5, Proposition 5.3, Appendix B.2–B.3.
1. Asymptotic Branch Step-Count Transfer to interpolationScale #
For 2 ≤ n', the Section 4.7 auxiliary dimension auxDim n' = r ^ (2 * Q) is strictly bounded
by n' itself (since D ≤ 2 ^ (λ / 128) ≤ 2 ^ (λ - 1) < n').
2. Outer Loop Certificates and Global Polynomial Workspace #
3. Public Branch Certificate, Polynomial Workspace, and Subexponential Runtime Bound #
Proposition 5.3 native branch certificate: uniform multi-tape Turing machine resource
certificate for evalBoundedIncidenceBranch on encBranchEvalInput
(OpenAI, The Exponential Time Hypothesis is False, §§5.3–5.5, Proposition 5.3,
pp. 37–39).
Instances For
Global polynomial workspace bound: boundedIncidenceBranchCertificate.space is
polynomially bounded in (encBranchEvalInput x).length across all raw BranchEvalInput
tuples x (including malformed or unfeasible inputs).
Proposition 5.3 subexponential branch runtime bound: for every degree bound k_deg,
polylogarithmic exponent β, and constant C_inc, above N₀_br, every feasible reduced branch
(n', sys') with sysMaxDegree sys' ≤ k_deg, sys'.length ≤ C_inc * n', and neutral-augmented
factor incidence and scope width at most C_inc * (Nat.clog 2 n') ^ β runs in
O^*(2^{O(interpolationScale n')}) native TM steps.
Instantiation of BranchPolylogTimeBound: boundedIncidenceBranchCertificate satisfies
Cslib.Complexity.BranchedSparseCNFCounter.BranchPolylogTimeBound for every N₀_branch.