Nonvanishing certificate for the input reduction
ProvedAPSCounting893.Main.input_reduction_nonvanishing_certificateLet have nonconstant coordinate polynomials, image-closure dimension , and maximum coordinate degree . The sampled coefficients comprise affine linear equations on the domain and linear combinations of the outputs. There is a nonzero complex polynomial in these coefficients with
such that every rational coefficient tuple with admits the selected rational affine parametrization, produces an image closure of dimension , and preserves APS in both directions. When , the sampled output coefficients are ignored, as in the source construction.
This algebraic exceptional-set certificate separates the geometric input-reduction argument from the finite-grid probability estimate in Proposition 3.14. It applies to every rational tuple, independently of the sampling grid.
Formalization Note The coefficients are enumerated by the canonical finite equivalence from the domain-equation and output-projection labels to an initial segment of the natural numbers. Successful parametrization, the composed projected polynomial map, its image dimension, and APS preservation are precisely the existing mission definitions.
import Definitions.Def_APSCounting893_Main_RandomReduction set_option autoImplicit false open APSCounting893.Main
theorem APSCounting893.Main.input_reduction_nonvanishing_certificate
(n m r : ℕ) (F : PolynomialMap ℂ n m)
(hnonconstant : ∀ i, 0 < (F i).totalDegree)
(hdim : RSCapacity.algDim (imageClosure F) = ((r : ℕ∞) : WithBot ℕ∞))
(hr : 1 ≤ r) (hrm : r < m) :
∃ p : MvPolynomial (Fin (Fintype.card (ReductionLabels n m r))) ℂ,
p ≠ 0 ∧ p.totalDegree ≤ 2 * (n-r) * inputDegree F ^ m + (r+1) * inputDegree F ^ r ∧
∀ s : ReductionLabels n m r → ℚ,
MvPolynomial.eval (fun j => (s ((Fintype.equivFin (ReductionLabels n m r)).symm j) : ℂ)) p ≠ 0 →
ReductionSucceeds F s := by sorry