H1 vanishes iff exact
ProvedMissingDataCohomology.DataComplex.h1_vanishes_iff_exactaether-cataloggeometry
Formal statement of MissingDataCohomology.DataComplex.h1_vanishes_iff_exact from the Aether Catalog (Geometry). The mathematical content is given by the Lean statement below; a human-readable write-up is pending.
theorem MissingDataCohomology.DataComplex.h1_vanishes_iff_exact(D : DataComplex 𝕜) :
Module.finrank 𝕜 D.H1 = 0 ↔ D.d1.ker = D.d0.range := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Geometry/MissingDataCohomology.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Geometry/RamseyTheory/MissingDataCohomology.lean
import Mathlib
import Definitions.Def_Geometry_FlagComplex
import Definitions.Def_Geometry_RamseyTheory_MissingDataCohomology
/-!
# Cohomological obstructions to patching incomplete data
Local observations and their overlap discrepancies form a two-step cochain
complex. Degree-zero cocycles are globally compatible observations, while the
first cohomology is the quotient of locally consistent discrepancies by those
arising from changing local observations. The results below give a rank formula,
a patchability criterion, and sharp boundary cases. They also identify the
nerve's flag condition as the precise passage from pairwise overlap consistency
to higher-order faces.
-- !-- Lab Notes -- !--
Hypothesis: the obstruction space is controlled by two independent rank losses,
not by the missing rate alone; pairwise overlap data determines all higher overlap
patterns exactly when the nerve is flag.
Experiment: zero differentials give obstruction dimension `dim C¹`, whereas a
surjective first coboundary gives dimension zero at the same ambient degree.
Analysis: rank-nullity and the quotient dimension formula show
`dim H¹ = dim C¹ - rank δ⁰ - rank δ¹`. Thus no universal function of a scalar
missing rate can determine obstruction dimension without assumptions on overlap
incidence and restriction-map ranks.
Critique: these are deterministic finite-dimensional results. They do not justify
a probabilistic asymptotic law, a likelihood interpretation, or comparisons with
particular statistical estimators; each requires an explicit generative model.
Synthesis: the rank formula separates the two sources of information loss, the
vanishing theorem characterizes exact patchability, and the imported clique-complex
construction links pairwise consistency to the topology of the data nerve.
-- !-- Lab Notes -- !--
-/
open Finset
open MissingDataCohomology
variable (𝕜 : Type*) [Field 𝕜]
attribute [instance] DataComplex.addC0 DataComplex.addC1 DataComplex.addC2
DataComplex.modC0 DataComplex.modC1 DataComplex.modC2 DataComplex.finC0
DataComplex.finC1 DataComplex.finC2
variable {𝕜}
/-
Every coboundary is a cocycle.
-/
/-
**Cohomological information-loss formula.** The obstruction dimension is
ambient overlap dimension minus the ranks of patch generation and consistency
checking.
-/
/-
Vanishing of first cohomology is equivalent to every cocycle being generated
by a degree-zero correction.
-/Formal statement
theorem MissingDataCohomology.DataComplex.h1_vanishes_iff_exact(D : DataComplex 𝕜) :
Module.finrank 𝕜 D.H1 = 0 ↔ D.d1.ker = D.d0.range := by sorrySource