Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

LIDT definitions: LDT / SelfImprovement / Defs

Definition
LIDT_MIPStarRE_LDT_SelfImprovement_Defs

by Dengnifer · Oct 27, 2026 · Mathlib 0df444a (Lean v4.33.1)

lidt

This bundle defines the canonical SDP witnesses and add-in-uuu data for self-improvement. It includes strict primal and dual witnesses, averaged point effects and dual slack, complementary slackness, selected point-polynomial pairs, sandwiched submeasurements, and the named error scales.

Principal definitions:

  • sdpDistinguishedPolynomial: The zero polynomial used as the distinguished SDP block.
  • sdpStrictPrimalWeight and sdpStrictPrimalSubMeas: The uniform strictly feasible primal witness with weight 1/(2∣P∣)1/(2|\mathcal P|)1/(2∣P∣).
  • sdpStrictDualWitness: The dual witness 2I2I2I.
  • averagedPointOperator: The averaged point effect Ag=EuAg(u)uA_g=\mathbb E_u A^u_{g(u)}Ag​=Eu​Ag(u)u​.
  • sdpDualSlackOperator: The dual slack Z−AgZ-A_gZ−Ag​.
  • sdpComplementarySlacknessEquation: The equation TgZ=TgAgT_gZ=T_gA_gTg​Z=Tg​Ag​.
  • AddInUSelection and addInUSelectionPairs: The selected point-indexed outcomes and polynomial pairs.
  • sandwichedPolynomialSubMeasAt, averagedSandwichedPolynomialSubMeas, and selfImprovementError: The sandwiched families and their error scales.

Formalization credit: Sirui Lu, Ruixuan Deng, Yanqiao Zhu, and Zhengfeng Ji (arXiv:2609.19814). Mathematical source: Zhengfeng Ji, Anand Natarajan, Thomas Vidick, John Wright, and Henry Yuen (arXiv:2009.12982).

Definition code
-- Generated from MIPStarRE/LDT/SelfImprovement/Defs.lean by skeleton subtraction.
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_AxisParallelLine
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DiagonalLine
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_Distribution
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DistributionAvg
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DistributionUniformSums
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_LinePolynomials
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_LowDegreePolynomial
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_MeasurementLift
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_OpFamily
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_OperatorExpectations
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_PMFAverages
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_PMFUniformAverages
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_ParametersBase
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_ParametersFiniteAnswers
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_QuantumState
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_SubMeasurementCore
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_SubMeasurementFamilies
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_TensorPlacement
import Definitions.Def_LIDT_MIPStarRE_LDT_ExpansionHypercubeGraph_Defs_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_GlobalVariance_Defs_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_GlobalVariance_Defs_Families
import Definitions.Def_LIDT_MIPStarRE_LDT_GlobalVariance_Defs_Operators
import Definitions.Def_LIDT_MIPStarRE_LDT_MainInductionStep_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_NaimarkCore
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Projectivization
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Statements
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Completion
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_CauchySchwarz
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_ComparisonCore
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_DistanceBounds
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichMain_Completeness
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_ApproxDelta
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_InnerProduct
import Definitions.Def_LIDT_MIPStarRE_LDT_Tactic_QuantumNonneg
import Definitions.Def_LIDT_MIPStarRE_LDT_Test_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_Test_StrategyCore
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_Basic
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_BlockDiagonal
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_NormalizedTrace
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_Order
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_TracePairing
import Definitions.Def_LIDT_MIPStarRE_Quantum_Measurement
import Mathlib
import Mathlib.Analysis.MeanInequalitiesPow
import Mathlib.Probability.ProbabilityMassFunction.Integrals
import Mathlib.Probability.ProbabilityMassFunction.Monad

namespace LIDT

namespace MIPStarRE.LDT.Preliminaries
end MIPStarRE.LDT.Preliminaries

/-!
# Section 9 — Definitions

This file exposes the paper's SDP witnesses, the `add-in-u` transfer identity,
and the non-projective/projective self-improvement outputs through explicit named
constructions and error terms.

## References

- `references/ldt-paper/self_improvement.tex`
-/

@[expose] public section

namespace MIPStarRE.LDT.SelfImprovement

open MIPStarRE.LDT
open MIPStarRE.Quantum
open MIPStarRE.LDT.ExpansionHypercubeGraph
open MIPStarRE.LDT.GlobalVariance
open MIPStarRE.LDT.MakingMeasurementsProjective
open scoped BigOperators MatrixOrder _root_.Matrix ComplexOrder

variable {ι : Type*} [Fintype ι] [DecidableEq ι]

/-- The zero polynomial used as the distinguished polynomial block in canonical
SDP completion arguments. -/
noncomputable def sdpDistinguishedPolynomial (params : Parameters) [FieldModel params.q] :
    Polynomial params :=
  ⟨0, by
    intro i
    simp [MvPolynomial.degreeOf_zero]⟩

/-- The paper's strict-feasibility weight `1 / (2 |\polyfunc{m}{q}{d}|)`. -/
noncomputable def sdpStrictPrimalWeight (params : Parameters)
    [FieldModel params.q] : Error :=
  1 / (2 * (Fintype.card (Polynomial params) : Error))

omit [Fintype ι] in
/-- Paper origin: `references/ldt-paper/self_improvement.tex:168-176`
(`\label{lem:sdp}` strict feasible primal witness
`T_g = (2 |\polyfunc{m}{q}{d}|)^{-1} I`).

The constant strict primal effects have total mass `(1/2)I`.  This is the
scalar identity behind the paper's strict feasible primal witness. -/
theorem sdpStrictPrimalConstantSum (params : Parameters)
    [FieldModel params.q] :
    ∑ _ : Polynomial params, sdpStrictPrimalWeight params •
        (1 : MIPStarRE.Quantum.Op ι) =
      ((1 / 2 : Error) • (1 : MIPStarRE.Quantum.Op ι)) := by
  have hdenom : (2 * (Fintype.card (Polynomial params) : Error)) ≠ 0 := by
    positivity
  calc
    ∑ g : Polynomial params, sdpStrictPrimalWeight params •
        (1 : MIPStarRE.Quantum.Op ι)
        = ((∑ g : Polynomial params, sdpStrictPrimalWeight params) : Error) •
            (1 : MIPStarRE.Quantum.Op ι) := by
              simpa using
                (Finset.sum_smul (s := Finset.univ)
                  (f := fun _ : Polynomial params => sdpStrictPrimalWeight params)
                  (x := (1 : MIPStarRE.Quantum.Op ι))).symm
    _ = ((Fintype.card (Polynomial params) : Error) *
          sdpStrictPrimalWeight params) •
          (1 : MIPStarRE.Quantum.Op ι) := by
            simp [Finset.sum_const, nsmul_eq_mul]
    _ = ((1 / 2 : Error) • (1 : MIPStarRE.Quantum.Op ι)) := by
          congr 1
          unfold sdpStrictPrimalWeight
          field_simp [hdenom]

/-- The paper's strict-feasible primal SDP witness
`T_g = (2 |\polyfunc{m}{q}{d}|)^{-1} I`. -/
noncomputable def sdpStrictPrimalSubMeas (params : Parameters)
    [FieldModel params.q] : SubMeas (Polynomial params) ι :=
  { outcome := fun _ => sdpStrictPrimalWeight params • (1 : MIPStarRE.Quantum.Op ι)
    total := ∑ g : Polynomial params,
      sdpStrictPrimalWeight params • (1 : MIPStarRE.Quantum.Op ι)
    outcome_pos := fun _ => smul_nonneg (by
      unfold sdpStrictPrimalWeight
      positivity) (Matrix.PosSemidef.one.nonneg : 0 ≤ (1 : MIPStarRE.Quantum.Op ι))
    sum_eq_total := rfl
    total_le_one := (le_of_eq (sdpStrictPrimalConstantSum (ι := ι) params)).trans (by
      simpa using smul_le_smul_of_nonneg_right
        (show (1 / 2 : Error) ≤ 1 by norm_num)
        (Matrix.PosSemidef.one.nonneg : 0 ≤ (1 : MIPStarRE.Quantum.Op ι))) }

/-- The paper's uniform strict-feasible primal witness has total mass
`(1 / 2) • I`. -/
@[simp] theorem sdpStrictPrimalSubMeas_total (params : Parameters)
    [FieldModel params.q] :
    (sdpStrictPrimalSubMeas (ι := ι) params).total =
      ((1 / 2 : Error) • (1 : MIPStarRE.Quantum.Op ι)) := by
  simpa [sdpStrictPrimalSubMeas] using
    sdpStrictPrimalConstantSum (ι := ι) params

/-- Paper origin: `references/ldt-paper/self_improvement.tex:168-176`
(`\label{lem:sdp}` strict feasible dual witness `Z = 2I`);
blueprint `\label{lem:sdp-uniform-feasible-witness}`.

The paper's strict-feasible dual SDP witness `Z = 2I`. -/
noncomputable def sdpStrictDualWitness : MIPStarRE.Quantum.Op ι :=
  (2 : Error) • (1 : MIPStarRE.Quantum.Op ι)

/-- The paper's strict-feasible dual witness `2I` is positive semidefinite. -/
@[simp] theorem sdpStrictDualWitness_nonneg {ι : Type*} [Finite ι] [DecidableEq ι] :
    0 ≤ (sdpStrictDualWitness (ι := ι)) := by
  let := Fintype.ofFinite ι
  unfold sdpStrictDualWitness
  exact smul_nonneg (by norm_num)
    (Matrix.PosSemidef.one.nonneg : 0 ≤ (1 : MIPStarRE.Quantum.Op ι))

/-- The paper's strict-feasible dual witness dominates the identity: `I ≤ 2I`. -/
theorem one_le_sdpStrictDualWitness {ι : Type*} [Finite ι] [DecidableEq ι] :
    (1 : MIPStarRE.Quantum.Op ι) ≤ sdpStrictDualWitness (ι := ι) := by
  let := Fintype.ofFinite ι
  simpa [sdpStrictDualWitness] using
    (smul_le_smul_of_nonneg_right
      (show (1 : Error) ≤ 2 by norm_num)
      (Matrix.PosSemidef.one.nonneg : 0 ≤ (1 : MIPStarRE.Quantum.Op ι)))

/-- The averaged point operator `A_g = E_u A^u_{g(u)}`. -/
noncomputable def averagedPointOperator (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι) (g : Polynomial params) : MIPStarRE.Quantum.Op ι :=
  averageOperatorOverDistribution (uniformDistribution (Point params))
    (pointConditionedOutcomeOperatorAtPolynomial params strategy g)

/-- The averaged point operator `A_g` is positive semidefinite. -/
theorem averagedPointOperator_nonneg (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι) (g : Polynomial params) :
    0 ≤ averagedPointOperator params strategy g := by
  unfold averagedPointOperator
  exact averageOperatorOverDistribution_nonneg (uniformDistribution (Point params))
    (pointConditionedOutcomeOperatorAtPolynomial params strategy g)
    (fun u => (strategy.pointMeasurement u).toSubMeas.outcome_pos (g u))







/-- The dual slack operator `Z - A_g`. -/
noncomputable def sdpDualSlackOperator (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι)
    (Z : MIPStarRE.Quantum.Op ι) (g : Polynomial params) : MIPStarRE.Quantum.Op ι :=
  Z - averagedPointOperator params strategy g

/-- Dual feasibility already implies that the dual operator is positive
semidefinite, since every averaged point operator `A_g` is positive. -/
theorem sdpDualPositive_of_dualFeasible (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι)
    (Z : MIPStarRE.Quantum.Op ι)
    (hdual :
      ∀ g : Polynomial params,
        0 ≤ sdpDualSlackOperator params strategy Z g) :
    0 ≤ Z := by
  -- Any polynomial would suffice here; the distinguished one is only a
  -- convenient fixed element of the finite polynomial type.
  let g0 : Polynomial params := sdpDistinguishedPolynomial params
  have hAg_nonneg : 0 ≤ averagedPointOperator params strategy g0 :=
    averagedPointOperator_nonneg params strategy g0
  have hAg_le_Z : averagedPointOperator params strategy g0 ≤ Z :=
    sub_nonneg.mp (by simpa [sdpDualSlackOperator] using hdual g0)
  exact hAg_nonneg.trans hAg_le_Z

/-- The complementary-slackness equation `T_g Z = T_g A_g`. -/
def sdpComplementarySlacknessEquation (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι)
    (T : SubMeas (Polynomial params) ι)
    (Z : MIPStarRE.Quantum.Op ι) (g : Polynomial params) : Prop :=
  T.outcome g * Z =
    T.outcome g * averagedPointOperator params strategy g

/-- A point-indexed selection of outcome/polynomial pairs used in `lem:add-in-u`. -/
abbrev AddInUSelection (params : Parameters) [FieldModel params.q] (Outcome : Type*) :=
  Point params → Set (Outcome × Polynomial params)

/-- The finite set of selected outcome/polynomial pairs at a point `u`. -/
noncomputable def addInUSelectionPairs {Outcome : Type*} [Fintype Outcome]
    (params : Parameters) [FieldModel params.q]
    (S : AddInUSelection params Outcome) (u : Point params) :
    Finset (Outcome × Polynomial params) :=
  open Classical in
    Finset.univ.filter (fun ah : Outcome × Polynomial params => ah ∈ S u)

/-- The pointwise sandwiched operator `H^u_h = A^u_{h(u)} T_h A^u_{h(u)}`. -/
noncomputable def sandwichedPolynomialOutcomeOperatorAt (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι)
    (T : SubMeas (Polynomial params) ι)
    (u : Point params) (h : Polynomial params) : MIPStarRE.Quantum.Op ι :=
  let Au := pointConditionedOutcomeOperatorAtPolynomial params strategy h u
  Au * (T.outcome h) * Au

/-- The sum of the pointwise sandwiched operators is bounded above by the identity. -/
private theorem sandwichedPolynomialOutcomeOperatorAt_sum_le_one (params : Parameters)
    [FieldModel params.q] (strategy : SymStrat params ι)
    (T : SubMeas (Polynomial params) ι) (u : Point params) :
    ∑ h : Polynomial params, sandwichedPolynomialOutcomeOperatorAt params strategy T u h ≤ 1 := by
  let Au := strategy.pointMeasurement u
  -- Regroup by evaluation value a = h(u).
  calc
    ∑ h : Polynomial params, sandwichedPolynomialOutcomeOperatorAt params strategy T u h
      = ∑ a : Fq params,
          ∑ h ∈ Finset.univ.filter (fun h : Polynomial params => h u = a),
            Au.toSubMeas.outcome a * T.outcome h * Au.toSubMeas.outcome a := by
          rw [show ∑ h : Polynomial params,
                sandwichedPolynomialOutcomeOperatorAt params strategy T u h =
              ∑ a : Fq params,
                ∑ h ∈ Finset.univ.filter (fun h : Polynomial params => h u = a),
                  sandwichedPolynomialOutcomeOperatorAt params strategy T u h from by
            exact polynomial_sum_fiberwise params u
              (sandwichedPolynomialOutcomeOperatorAt params strategy T u)]
          refine Finset.sum_congr rfl ?_
          intro a _
          refine Finset.sum_congr rfl ?_
          intro h hh
          simp only [sandwichedPolynomialOutcomeOperatorAt,
            pointConditionedOutcomeOperatorAtPolynomial]
          simp only [Finset.mem_filter, Finset.mem_univ, true_and] at hh
          simp [Au, hh]
    _ = ∑ a : Fq params,
          Au.toSubMeas.outcome a *
            (∑ h ∈ Finset.univ.filter (fun h : Polynomial params => h u = a),
              T.outcome h) *
            Au.toSubMeas.outcome a := by
          refine Finset.sum_congr rfl ?_
          intro a _
          rw [← Matrix.sum_mul, ← Matrix.mul_sum]
    _ ≤ ∑ a : Fq params, Au.toSubMeas.outcome a := by
          refine Finset.sum_le_sum ?_
          intro a _
          have hfilt_le_one : ∑ h ∈ Finset.univ.filter
              (fun h : Polynomial params => h u = a), T.outcome h ≤ 1 :=
            calc
              ∑ h ∈ Finset.univ.filter (fun h : Polynomial params => h u = a), T.outcome h
                ≤ ∑ h : Polynomial params, T.outcome h :=
                  Finset.sum_le_sum_of_subset_of_nonneg
                    (Finset.filter_subset _ _) (fun h _ _ => T.outcome_pos h)
              _ = T.total := T.sum_eq_total
              _ ≤ 1 := T.total_le_one
          simpa [Au.proj a] using
            IsSelfAdjoint.conjugate_le_conjugate
              (c := Au.toSubMeas.outcome a) hfilt_le_one (Au.outcome_hermitian a)
    _ = Au.toSubMeas.total := by
          rw [Au.toSubMeas.sum_eq_total]
    _ = 1 := by
          simpa using Au.total_eq_one

/-- The pointwise sandwiched submeasurement `H^u = {H^u_h}`. -/
noncomputable def sandwichedPolynomialSubMeasAt (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι)
    (T : SubMeas (Polynomial params) ι) (u : Point params) :
    SubMeas (Polynomial params) ι :=
  { outcome := sandwichedPolynomialOutcomeOperatorAt params strategy T u
    total := ∑ h : Polynomial params,
      sandwichedPolynomialOutcomeOperatorAt params strategy T u h
    outcome_pos := fun h => by
      simp only [sandwichedPolynomialOutcomeOperatorAt, pointConditionedOutcomeOperatorAtPolynomial]
      exact IsSelfAdjoint.conjugate_nonneg (T.outcome_pos h)
        (SubMeas.outcome_hermitian (strategy.pointMeasurement u).toSubMeas (h u))
    sum_eq_total := rfl
    total_le_one := by
      exact sandwichedPolynomialOutcomeOperatorAt_sum_le_one params strategy T u }

/-- The average of the total pointwise sandwiched operators is bounded by the identity. -/
private theorem averagedSandwichedPolynomialSubMeas_total_le_one (params : Parameters)
    [FieldModel params.q] (strategy : SymStrat params ι)
    (T : SubMeas (Polynomial params) ι) :
    ∑ h : Polynomial params,
      averageOperatorOverDistribution (uniformDistribution (Point params))
        (fun u => sandwichedPolynomialOutcomeOperatorAt params strategy T u h) ≤ 1 := by
  let 𝒟 := uniformDistribution (Point params)
  calc
    ∑ h : Polynomial params,
        averageOperatorOverDistribution 𝒟
          (fun u => sandwichedPolynomialOutcomeOperatorAt params strategy T u h)
      = averageOperatorOverDistribution 𝒟
          (fun u => ∑ h : Polynomial params,
            sandwichedPolynomialOutcomeOperatorAt params strategy T u h) := by
            exact (averageOperatorOverDistribution_sum 𝒟
              (fun u h => sandwichedPolynomialOutcomeOperatorAt params strategy T u h)).symm
    _ = averageOperatorOverDistribution 𝒟
          (fun u => (sandwichedPolynomialSubMeasAt params strategy T u).total) := by
            exact averageOperatorOverDistribution_congr 𝒟 _ _ fun u => by
              simp [sandwichedPolynomialSubMeasAt]
    _ ≤ 1 := by
          simpa [𝒟] using
            averageOperatorOverDistribution_uniform_le_one
              (fun u => (sandwichedPolynomialSubMeasAt params strategy T u).total)
              (fun u => (sandwichedPolynomialSubMeasAt params strategy T u).total_le_one)

/-- The averaged sandwiched submeasurement `H_h = E_u H^u_h`. -/
noncomputable def averagedSandwichedPolynomialSubMeas (params : Parameters)
    [FieldModel params.q]
    (strategy : SymStrat params ι)
    (T : SubMeas (Polynomial params) ι) : SubMeas (Polynomial params) ι :=
  { outcome := fun h =>
      averageOperatorOverDistribution (uniformDistribution (Point params))
        (fun u => sandwichedPolynomialOutcomeOperatorAt params strategy T u h)
    total := ∑ h : Polynomial params,
      averageOperatorOverDistribution (uniformDistribution (Point params))
        (fun u => sandwichedPolynomialOutcomeOperatorAt params strategy T u h)
    outcome_pos := fun h => averageOperatorOverDistribution_nonneg
      (uniformDistribution (Point params))
      (fun u => sandwichedPolynomialOutcomeOperatorAt params strategy T u h)
      (fun u => (sandwichedPolynomialSubMeasAt params strategy T u).outcome_pos h)
    sum_eq_total := rfl
    total_le_one := by
      exact averagedSandwichedPolynomialSubMeas_total_le_one params strategy T }

/-- The variance error entering `lem:add-in-u`. -/
noncomputable def selfImprovementVarianceError (params : Parameters)
    [FieldModel params.q]
    (eps delta : Error) : Error :=
  globalVarianceOfPointsError params eps delta

/-- The error term in `lem:add-in-u`. -/
noncomputable def addInUError (params : Parameters)
    [FieldModel params.q]
    (eps delta : Error) : Error :=
  4 * Real.rpow (selfImprovementVarianceError params eps delta) (1 / (2 : Error))

/-- The quantitative error from `lem:self-improvement-helper`. -/
noncomputable def selfImprovementHelperError (params : Parameters)
    [FieldModel params.q]
    (eps delta : Error) : Error :=
  100 * (params.m : Error) *
    (Real.rpow eps (1 / (2 : Error)) +
      Real.rpow delta (1 / (2 : Error)) +
      Real.rpow (((params.d : Error) / (params.q : Error))) (1 / (2 : Error)))

/-- The orthogonalization error applied to the helper output. -/
noncomputable def selfImprovementOrthogonalizationError (params : Parameters)
    [FieldModel params.q]
    (eps delta : Error) : Error :=
  orthonormalizationError (selfImprovementHelperError params eps delta)

/-- The postprocessed error after projecting the helper output. -/
noncomputable def selfImprovementDataProcessingError (params : Parameters)
    [FieldModel params.q]
    (eps delta : Error) : Error :=
  8 * selfImprovementHelperError params eps delta +
    8 * Real.rpow (selfImprovementOrthogonalizationError params eps delta)
      (1 / (2 : Error))

/-- The quantitative error from `thm:self-improvement`. -/
noncomputable def selfImprovementError (params : Parameters)
    [FieldModel params.q]
    (eps delta : Error) : Error :=
  MainInductionStep.selfImprovementInInductionError params eps delta 0

end MIPStarRE.LDT.SelfImprovement

end

end LIDT
Source
https://github.com/LionSR/MIPStarRE/blob/f382c801026f53a9efc260fdee707cb35e0adb15/MIPStarRE/LDT/SelfImprovement/Defs.lean#L32-L384

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me