The concrete LLL reducer preserves the original column lattice exactly
ProvedInfoLossLAR.WorstCase.NativeLLL.basis_lattice_eqformal-verificationlatticeslll
For every square integer matrix B, the column matrix basis B produced by the concrete logarithmic-budget LLL reducer generates exactly the same set of Euclidean lattice vectors as B. The statement is an equality of the original lattice sets and includes singular inputs. The row-basis implementation and the column-basis interface are connected by the explicitly defined transpose conversion; no replacement lattice or approximate membership relation is used.
Preamble
import Definitions.Def_InfoLossLAR_IntegerBasis_Core import Definitions.Def_InfoLossLAR_WorstCase_LatticeProblems import Definitions.Def_LARLLL_CurrentData import Definitions.Def_QuantumMaxCutFiniteNumericRuntime import Init import Init.Data.List.Sort.Basic import Init.Grind.Ring.Field import Lean import Mathlib import Mathlib.Algebra.Module.ZLattice.Basic import Mathlib.Algebra.Module.ZLattice.Covolume import Mathlib.Algebra.Module.ZLattice.Summable import Mathlib.Algebra.MvPolynomial.Variables import Mathlib.Algebra.Polynomial.Eval.Defs import Mathlib.Analysis.Distribution.SchwartzSpace.Basic import Mathlib.Analysis.Distribution.SchwartzSpace.Fourier import Mathlib.Analysis.Fourier.AddCircleMulti import Mathlib.Analysis.Fourier.FourierTransform import Mathlib.Analysis.Fourier.PoissonSummation import Mathlib.Analysis.InnerProductSpace.GramMatrix import Mathlib.Analysis.InnerProductSpace.GramSchmidtOrtho import Mathlib.Analysis.InnerProductSpace.PiL2 import Mathlib.Analysis.Matrix.LDL import Mathlib.Analysis.Normed.Ring.InfiniteSum import Mathlib.Analysis.SpecialFunctions.Exponential import Mathlib.Analysis.SpecialFunctions.Gaussian.PoissonSummation import Mathlib.Computability.TuringMachine.Computable import Mathlib.Data.Fin.Tuple.Basic import Mathlib.Data.Fin.Tuple.Reflection import Mathlib.Data.Finset.Lattice.Fold import Mathlib.Data.Finset.Sort import Mathlib.Data.Fintype.BigOperators import Mathlib.Data.List.DropRight import Mathlib.Data.List.MinMax import Mathlib.Data.Nat.Factorial.BigOperators import Mathlib.Data.Nat.Log import Mathlib.Data.Rat.Floor import Mathlib.Data.Vector.Basic import Mathlib.GroupTheory.Perm.Fin import Mathlib.LinearAlgebra.BilinearForm.DualLattice import Mathlib.LinearAlgebra.Dimension.Constructions import Mathlib.LinearAlgebra.Finsupp.LinearCombination import Mathlib.LinearAlgebra.Matrix.AbsoluteValue import Mathlib.LinearAlgebra.Matrix.Adjugate import Mathlib.LinearAlgebra.Matrix.Determinant.Basic import Mathlib.LinearAlgebra.Matrix.PosDef import Mathlib.LinearAlgebra.Matrix.Rank import Mathlib.LinearAlgebra.Matrix.Reindex import Mathlib.LinearAlgebra.Matrix.Swap import Mathlib.LinearAlgebra.Matrix.ToLin import Mathlib.LinearAlgebra.Matrix.ToLinearEquiv import Mathlib.LinearAlgebra.Matrix.Transvection import Mathlib.Logic.Equiv.Fin.Basic import Mathlib.MeasureTheory.Function.LpSeminorm.LpNorm import Mathlib.MeasureTheory.Measure.Haar.InnerProductSpace import Mathlib.MeasureTheory.Measure.Haar.Unique import Mathlib.Order.Hom.Set import Mathlib.RingTheory.MvPolynomial.Basic import Mathlib.RingTheory.MvPolynomial.Homogeneous import Mathlib.RingTheory.MvPolynomial.Symmetric.NewtonIdentities import Mathlib.Tactic import Std import Std.Data.ExtTreeMap.Lemmas set_option autoImplicit false set_option maxRecDepth 8192
Formal statement
theorem _root_.InfoLossLAR.WorstCase.NativeLLL.basis_lattice_eq : ∀ {n : Nat} (B : Matrix.{0, 0, 0} (Fin n) (Fin n) Int),
@Eq.{1} (Set.{0} (EuclideanSpace.{0, 0} Real (Fin n)))
(@InfoLossLAR.WorstCase.lattice n (@InfoLossLAR.WorstCase.NativeLLL.basis n B)) (@InfoLossLAR.WorstCase.lattice n B) := by sorrySource
Existing checked theorem InfoLossLAR.WorstCase.NativeLLL.basis_lattice_eq in Solutions.RegevBridgeNativeLLLOriginalBasis.lean. The submitted declaration and its supporting proofs are extracted from the checked Lean source using the kernel dependency graph and Lean syntax positions. The vendor mathematics uses the frozen leanprover/hex-lll dependency family; original Apache-2.0 license and copyright notices are retained in the proof sources. This submission adds no new mathematical claim. Existing checked Lean theorem from Solutions.RegevBridgeNativeLLLOriginalBasis, original source lines 44–67. Original proof commands are selected using the official Lean declaration/sketch and syntax oracles, with only declared-name/reference remapping. Recursive self references, when renamed, are recovered from matching original Lean local-fvar identities. The original theorem and all inlined helper types are kernel checked; the complete proof and all its generated declarations use only the standard three axioms. Apache-2.0 notices and source provenance are preserved.