GaussianAnalysisRocq library documentation
Main resultsGlossary
On this page
GaussianApplicationsapplicationsPerceptronSignStatement

Sign and Gaussian patterns of the Ising perceptron: the statement

GaussianApplications.PerceptronSignStatementapplications/PerceptronSignStatement.v59 lines1 definitionsource
Imports 3 library modules and 20 external

External: HB.structures, mathcomp.classical.all_ssreflect_compat, mathcomp.algebra.algebraic_hierarchy.ssralg, mathcomp.algebra.numeric_hierarchy.ssrnum, mathcomp.algebra.matrix, mathcomp.classical.boolp, mathcomp.classical.classical_sets, mathcomp.classical.functions, mathcomp.reals.reals, mathcomp.analysis.ereal, mathcomp.analysis.topology_theory.topology, mathcomp.analysis.normedtype_theory.normedtype, mathcomp.analysis.sequences, mathcomp.analysis.derive, mathcomp.analysis.exp, mathcomp.analysis.measure_theory.measure, mathcomp.analysis.lebesgue_measure, mathcomp.analysis.lebesgue_integral_theory.lebesgue_integral, mathcomp.analysis.measurable_realfun, mathcomp.analysis.probability_theory.probability

Imported by 1

Target T6 of phase 12 (AGENTS.md, section Phase 12; the plan notes/phase12-plan.md, section 3). The statement of a Rocq formalization of the credited argument of the first paragraph of the proof of Theorem ext:sign of 'The free energy of the Ising random perceptron' (family 222 of the OpenAI math collection, revision 1581d4cbd, sections/extensions.tex): the coordinate-replacement form of the Lindeberg principle, with a third-derivative bound C N^(-5/2) in one pattern entry. The collection has no Lean version of it. This file holds the statement only; its definitions are those of PerceptronSignDefinitions.v.

perceptron_sign_statement, for every R and all bounds B0, B1, B2, B3: there is C > 0 such that for every potential f of class C^3 with |f| <= B0, |f'| <= B1, |f''| <= B2 and |f'''| <= B3 (bounded_C3), every number of patterns M and every N >= 1,

(a) g |-> p_N(vec_mx g, f) is integrable against gauss (M * N), so that gauss_pressure M N f is the expectation of the pressure under Gaussian patterns;

(b) |E_sign p_N - E_Gauss p_N| <= C M N^(-3/2), the power being powR.

Choices, for the reviewer.

  • The constant C depends only on the four bounds, uniformly in M and N, as the user asked (formally also on R).
  • The class of potentials: f of class C^3 with f, f', f'' and f''' bounded, which contains the manuscript's C_c^infinity, as the user accepted. The extension to bounded Borel potentials whose discontinuities form a Gaussian-null set, the rest of Theorem ext:sign, is not part of the statement; for arbitrary bounded Borel potentials the comparison fails, as the manuscript remarks.
  • M is arbitrary, not M_N = floor (alpha N): with M_N N entries, (b) is the manuscript's error O_f(N^(-1/2)).

Declarations1 exported

lemma, theoremdefinitioninstancelocal
Type
Prop
Definition
Definition perceptron_sign_statement : Prop :=
  forall (R : realType) (B0 B1 B2 B3 : R),
  exists C : R, 0 < C /\
  forall (f : R -> R) (M N : nat), bounded_C3 f B0 B1 B2 B3 -> (0 < N)%N ->
    (gauss (M * N)).-integrable [set: 'rV[R]_(M * N)]
      (EFin \o (fun g : 'rV[R]_(M * N) => perceptron_pressure f (vec_mx g))) /\
    `|sign_pressure M N f - gauss_pressure M N f| <=
      C * M%:R * N%:R `^ (- (3 / 2)).
Used by 1 declaration in 1 file
Generated from commit 9046883 (tag phase12) of the ellipsoid fitting repository. Signatures are printed by Rocq 9.1.1; links in the source statements come from the compiled .glob files. The source links open the files on GitHub at this commit, in the now-public repository.