Stat.Nonparametric.Moment­Problems

Moment-problem substrate for nonparametric statistics: raw-moment algebra, L² projection residuals, constrained score programs, and sharp bounded-outcome residual envelopes.

Residual­Quadratic 23 core · 18 supporting · 3 submodules Residual-quadratic moment problems: moment-coordinate formulas, measure-level L² projection residuals, and projection residual orthogonality identities. Bounded­Outcome­Envelope 14 core · 26 supporting · 5 submodules Sharp bounded-outcome residual envelopes: quartic root selection, admissible laws, upper bounds, and extremal three-point attainment. Finite­Moment­Near­Gaussian­Perturbation 8 core · 1 supporting · 6 submodules Finite-moment Gaussian perturbations: bounded density tilts that preserve any chosen finite segment of Gaussian moments while remaining non-Gaussian, arbitrarily close in total variation, and Carleman-certified. Score­Program 4 core · 3 supporting · 1 submodule Dual score-program tools for moment problems: projection residual functions, feasible score directions, quadratic objectives, and minimum-norm identities.
Raw­Moment 2 core · 0 supporting This module fixes the single raw-moment notation shared by the MomentProblems developments: the order-k raw moment of a measure on the real line is the Bochner integral of the k-th power. ★ rawMoment_eq_integral

Raw moments of a measure on the real line

This module fixes the single raw-moment notation shared by the MomentProblems developments: the order-k raw moment of a measure on the real line is the Bochner integral of the k-th power. Both the finite-moment near-Gaussian perturbation construction and the measure bridge for the quadratic projection residual are stated in terms of it.

def rawMoment reviewed
Causalean.Stat.MomentProblems

Given a real measure and a nonnegative integer order, the raw moment is given by integrating the corresponding power. This is the real-valued Bochner integral, which is zero when the power is not integrable, so it is the moment only when the moment exists.

Definition (Lean source)
ν :
k :
rawMoment ν k :
∫ x, x ^ k ∂ν
Causalean.Stat.MomentProblems.rawMoment · Causalean/Stat/Nonparametric/MomentProblems/RawMoment.lean:21
theorem rawMoment_eq_integral reviewed
Causalean.Stat.MomentProblems

For a real measure at a nonnegative integer order, the raw-moment notation equals the integral of the corresponding power.

Formal statement
ν :
k :
rawMoment ν k = ∫ x, x ^ k ∂ν
Proof (Lean source)
@[simp] theorem rawMoment_eq_integral (ν : Measure ℝ) (k : ℕ) : rawMoment ν k = ∫ x, x ^ k ∂ν := rfl
Causalean.Stat.MomentProblems.rawMoment_eq_integral · Causalean/Stat/Nonparametric/MomentProblems/RawMoment.lean:28 · uses rawMoment
Cumulant 3 core · 0 supporting This module provides the general moment-to-cumulant coordinates used throughout the moment-problem layer: the joint cumulant of a pair of real random variables at a given bidegree (defined by the set-partition Möbius for

Cumulants of real random variables, and the Gaussian-law predicate

This module provides the general moment-to-cumulant coordinates used throughout the moment-problem layer: the joint cumulant of a pair of real random variables at a given bidegree (defined by the set-partition Möbius formula on mixed moments), the univariate cumulant of a single real random variable obtained by specializing it, and the predicate "this law on the real line is a Gaussian law".

These are the plain textbook objects: the second cumulant is the variance, the third is the centered third moment, the fourth is the excess kurtosis, and a Gaussian law is exactly a law whose cumulants above order two all vanish. The Gaussian predicate here is the distributional one — being some gaussianReal with a mean and a variance, degenerate Diracs included — and is the form in which non-Gaussianity assumptions (as in linear non-Gaussian causal discovery) are stated.

def IsGaussianLaw reviewed
Causalean.Stat.MomentProblems

For a measure on the real line, the Gaussian-law property holds exactly when there exist a real mean and a nonnegative variance parameter such that the measure is the associated normal distribution. The variance parameter may be zero, so point masses count as degenerate Gaussian laws.

Definition (Lean source)
ν :
IsGaussianLaw ν :
Prop
∃ (mean : ℝ) (v : ℝ≥0), ν = gaussianReal mean v
Causalean.Stat.MomentProblems.IsGaussianLaw · Causalean/Stat/Nonparametric/MomentProblems/Cumulant.lean:34
def jointCumulant reviewed
Causalean.Stat.MomentProblems

Given a sample space equipped with a σ-algebra, a measure on that space, two real-valued random variables, and two nonnegative integer orders, the joint cumulant at the specified bidegree is the sum over every partition of the combined slots, with the first pp slots assigned to the first variable and the remaining qq slots assigned to the second, of a signed factorial weight times the product, across its blocks, of the mixed moment whose exponents equal that block's counts of first- and second-variable slots.

Definition (Lean source)
Ω :
Type u_1
shared
μ :
X Y :
Ω → ℝ
p q :
jointCumulant μ X Y p q :
∑ π : Finpartition (Finset.univ : Finset (Fin (p + q))), (-1 : ℝ) ^ (π.parts.card - 1) * (factorial (π.parts.card - 1) : ℝ) * ∏ B ∈ π.parts, ∫ ω, (X ω) ^ (B.filter (fun i => i.val < p)).card * (Y ω) ^ (B.filter (fun i => p ≤ i.val)).card ∂μ
Causalean.Stat.MomentProblems.jointCumulant · Causalean/Stat/Nonparametric/MomentProblems/Cumulant.lean:45
def sourceCumulant reviewed
Causalean.Stat.MomentProblems

Given a sample space equipped with a σ-algebra, a measure on that space, a real-valued random variable, and a nonnegative integer order, the cumulant of that order is the joint cumulant obtained by placing that same variable in every slot.

Definition (Lean source)
Ω :
Type u_1
shared
μ :
S :
Ω → ℝ
r :
sourceCumulant μ S r :
jointCumulant μ S S r 0
Causalean.Stat.MomentProblems.sourceCumulant · Causalean/Stat/Nonparametric/MomentProblems/Cumulant.lean:67
Moment­Cumulant­Inversion 6 core · 8 supporting Cumulants are polynomials in the moments, and the polynomial of order r involves the moment of order r linearly (with coefficient one) plus a remainder built only from strictly lower moments. ★ sourceCumulant_eq_cumFromMom★ cumFromMom_momFromCum★ continuous_momFromCum

The triangular moment ↔ cumulant change of coordinates

Cumulants are polynomials in the moments, and the polynomial of order r involves the moment of order r linearly (with coefficient one) plus a remainder built only from strictly lower moments. That triangularity means the change of coordinates inverts: prescribing a cumulant sequence determines a moment sequence, recursively and uniquely, among sequences normalized to have total mass one and mean zero.

This module sets up both directions on abstract sequences of reals — no measure is needed to state them — and connects the forward direction to the measure-theoretic cumulant of a real random variable. It also records the two structural facts the inversion is used for downstream: each reconstructed moment depends only on the prescribed cumulants up to the same order (locality), and depends continuously on them.

def cumFromMom reviewed
Causalean.Stat.MomentProblems

For a nonnegative order and a real moment sequence, the cumulant read from that moment sequence is the signed, factorial-weighted sum over all partitions of a set of that many elements, with each partition contributing the product of the moments indexed by its block sizes.

Definition (Lean source)
r :
m :
ℕ → ℝ
cumFromMom r m :
∑ π : Finpartition (Finset.univ : Finset (Fin r)), (-1 : ℝ) ^ (π.parts.card - 1) * (factorial (π.parts.card - 1) : ℝ) * ∏ B ∈ π.parts, m B.card
def restFromMom reviewed
Causalean.Stat.MomentProblems

For a nonnegative order and a real moment sequence, the lower-order remainder of the cumulant formula is the same partition sum with the sole one-block partition omitted; hence it uses only moments of order strictly below that order.

Definition (Lean source)
r :
m :
ℕ → ℝ
restFromMom r m :
∑ π ∈ Finset.univ.filter (fun π : Finpartition (Finset.univ : Finset (Fin r)) => π.parts.card ≠ 1), (-1 : ℝ) ^ (π.parts.card - 1) * (factorial (π.parts.card - 1) : ℝ) * ∏ B ∈ π.parts, m B.card
theorem sourceCumulant_eq_cumFromMom reviewed
Causalean.Stat.MomentProblems

For a real random variable with law ν and any order r, the order-r cumulant of ν equals the abstract combinatorial cumulant formula evaluated at ν's own raw-moment sequence: the measure-theoretic and combinatorial definitions agree.

Formal statement
ν :
r :
sourceCumulant ν (id : ℝ → ℝ) r = cumFromMom r (fun k => ∫ t, t ^ k ∂ν)
Proof (Lean source)
theorem sourceCumulant_eq_cumFromMom (ν : Measure ℝ) (r : ℕ) : sourceCumulant ν (id : ℝ → ℝ) r = cumFromMom r (fun k => ∫ t, t ^ k ∂ν) := by unfold sourceCumulant jointCumulant cumFromMom apply Finset.sum_congr rfl intro π _ congr 1 apply Finset.prod_congr rfl intro B hB have hfirst : B.filter (fun i => i.val < r) = B := Finset.filter_true_of_mem (fun i _ => i.isLt) have hsecond : B.filter (fun i => r ≤ i.val) = ∅ := Finset.filter_false_of_mem (fun i _ hi => (not_le_of_gt i.isLt) hi) rw [hsecond] simp
Causalean.Stat.MomentProblems.sourceCumulant_eq_cumFromMom · Causalean/Stat/Nonparametric/MomentProblems/MomentCumulantInversion.lean:80 · uses cumFromMom , sourceCumulant
def momFromCum reviewed
Causalean.Stat.MomentProblems

For a real sequence of prescribed cumulants, the reconstructed moment sequence is defined by giving its zeroth moment the value one, giving its first moment the value zero, and at every order at least two, subtracting from the prescribed cumulant the lower-order remainder computed from the already reconstructed moments.

Definition (Lean source)
c :
ℕ → ℝ
momFromCum c :
ℕ → ℝ
clause 1
| 0 => 1
clause 2
| 1 => 0
clause 3
| succ (succ r)
=> c (r + 2)
- restFromMom (r + 2) (fun k => if _h : k < r + 2 then momFromCum c k else 0) termination_by n
=> n decreasing_by omega
theorem cumFromMom_momFromCum reviewed
Causalean.Stat.MomentProblems

Correctness of the inversion. Reading the cumulants back off the reconstructed moment sequence returns the prescribed cumulants, at every order r at least two.

Formal statement
c :
ℕ → ℝ
r :
hr :
2 ≤ r
Proof (Lean source)
theorem cumFromMom_momFromCum (c : ℕ → ℝ) (r : ℕ) (hr : 2 ≤ r) : cumFromMom r (momFromCum c) = c r := by rw [cumFromMom_eq r (by omega), momFromCum_succ c r hr] ring
theorem continuous_momFromCum reviewed
Causalean.Stat.MomentProblems

Continuity of the inversion. At any fixed order r, the reconstructed moment of that order is a continuous function of the prescribed cumulant sequence, so small perturbations of the target cumulants move the moments only slightly — the key to the openness arguments that use this inversion.

Formal statement
r :
Continuous (fun c : ℕ → ℝ => momFromCum c r)
Proof (Lean source)
@[fun_prop] theorem continuous_momFromCum (r : ℕ) : Continuous (fun c : ℕ → ℝ => momFromCum c r) := by induction r using Nat.strong_induction_on with | h r ih => by_cases hr : 2 ≤ r · simp_rw [momFromCum_succ _ r hr] apply (continuous_apply r).sub unfold restFromMom apply continuous_finset_sum intro π _ apply continuous_const.mul apply continuous_finset_prod intro B hB exact ih B.card (block_card_lt_of_card_ne_one π ((Finset.mem_filter.mp ‹π ∈ Finset.univ.filter _›).2) hB) · interval_cases r · simpa only [momFromCum_zero] using (continuous_const : Continuous (fun _ : ℕ → ℝ => (1 : ℝ))) · simpa only [momFromCum_one] using (continuous_const : Continuous (fun _ : ℕ → ℝ => (0 : ℝ)))
Causalean.Stat.MomentProblems.continuous_momFromCum · Causalean/Stat/Nonparametric/MomentProblems/MomentCumulantInversion.lean:221 · uses momFromCum
8 supporting declarations (lemmas, instances)
Atomic­Law 4 core · 4 supporting This module builds the finitely supported probability laws used as explicit witnesses in truncated moment problems: a finite list of distinct real atom locations carrying a matching list of weights. ★ isProbabilityMeasure_atomicLaw★ not_isGaussianLaw_atomicLaw

Finite atomic laws on the real line

This module builds the finitely supported probability laws used as explicit witnesses in truncated moment problems: a finite list of distinct real atom locations carrying a matching list of weights. It records how such a law integrates an arbitrary function (a weighted sum over the atoms), when it is a probability measure, how much mass it puts on each atom, that all of its moments are finite (the support is bounded), and that it is never Gaussian as soon as two atoms carry strictly positive mass.

The last fact is what makes finite atomic laws usable as non-Gaussian witnesses: a nondegenerate normal law has no atoms at all, and a degenerate one is a single point mass, so two distinct atoms of positive mass rule out both cases.

def atomicWeight reviewed
Causalean.Stat.MomentProblems

For a natural number of atoms, a list of their real locations, and a matching list of real weights, the weight assigned to a real point is the sum of the weights of precisely those atoms located at that point.

Definition (Lean source)
n :
x :
Fin n → ℝ
p :
Fin n → ℝ
atomicWeight n x p :
ℝ → ℝ
fun t ↦ ∑ i : Fin n, if x i = t then p i else 0
Causalean.Stat.MomentProblems.atomicWeight · Causalean/Stat/Nonparametric/MomentProblems/AtomicLaw.lean:30
def atomicLaw reviewed
Causalean.Stat.MomentProblems

For a natural number of atoms, a list of their real locations, and a matching list of real weights, the finite atomic law is the measure that puts each weight as point mass at its corresponding location.

Definition (Lean source)
n :
x :
Fin n → ℝ
p :
Fin n → ℝ
atomicLaw n x p :
Causalean.Stat.MomentProblems.atomicLaw · Causalean/Stat/Nonparametric/MomentProblems/AtomicLaw.lean:36
theorem isProbabilityMeasure_atomicLaw reviewed
Causalean.Stat.MomentProblems

A finite atomic law on the real line, built from n pairwise distinct atom locations carrying nonnegative weights that sum to one, is a probability measure.

Formal statement
n :
Fin n → ℝ
Fin n → ℝ
hx :
hp :
∀ i, 0 ≤ p i
hsum :
∑ i, p i = 1
Proof (Lean source)
theorem isProbabilityMeasure_atomicLaw {n : ℕ} {x : Fin n → ℝ} {p : Fin n → ℝ} (hx : Injective x) (hp : ∀ i, 0 ≤ p i) (hsum : ∑ i, p i = 1) : IsProbabilityMeasure (atomicLaw n x p) := by classical apply isProbabilityMeasure_discreteMeasure · intro t ht obtain ⟨i, hi, rfl⟩ := Finset.mem_image.mp ht rw [atomicWeight_apply hx] exact hp i · rw [Finset.sum_image hx.injOn] simpa only [atomicWeight_apply hx] using hsum
Causalean.Stat.MomentProblems.isProbabilityMeasure_atomicLaw · Causalean/Stat/Nonparametric/MomentProblems/AtomicLaw.lean:68 · uses atomicLaw
theorem not_isGaussianLaw_atomicLaw reviewed
Causalean.Stat.MomentProblems

A finite atomic probability law with at least two locations that are pairwise distinct and each carry strictly positive mass is not a Gaussian law: a normal law with positive variance has no point masses at all, and one with zero variance is a single point mass.

Formal statement
n :
Fin n → ℝ
Fin n → ℝ
hx :
hp :
∀ i, 0 < p i
hn :
2 ≤ n
Proof (Lean source)
theorem not_isGaussianLaw_atomicLaw {n : ℕ} {x : Fin n → ℝ} {p : Fin n → ℝ} (hx : Injective x) (hp : ∀ i, 0 < p i) (hn : 2 ≤ n) : ¬ IsGaussianLaw (atomicLaw n x p) := by classical intro hgauss obtain ⟨mean, v, hv⟩ := hgauss let i₀ : Fin n := ⟨0, by omega⟩ let i₁ : Fin n := ⟨1, by omega⟩ have hi₀₁ : i₀ ≠ i₁ := by intro h have := congrArg val h simp [i₀, i₁] at this have hx₀₁ : x i₀ ≠ x i₁ := fun h ↦ hi₀₁ (hx h) have hpnonneg : ∀ i, 0 ≤ p i := fun i ↦ (hp i).le by_cases hvzero : v = 0 · have hdirac : atomicLaw n x p = Measure.dirac mean := by rw [hv, hvzero, gaussianReal_zero_var] have hmass₀ := atomicLaw_singleton hx hpnonneg i₀ have hmass₁ := atomicLaw_singleton hx hpnonneg i₁ rw [hdirac] at hmass₀ hmass₁ have hmean₀ : mean = x i₀ := by by_contra hne have hzero : Measure.dirac mean {x i₀} = 0 := by simp [Measure.dirac_apply' mean (MeasurableSet.singleton (x i₀)), hne] exact ENNReal.ofReal_ne_zero_iff.mpr (hp i₀) (hmass₀.symm.trans hzero) have hmean₁ : mean = x i₁ := by by_contra hne have hzero : Measure.dirac mean {x i₁} = 0 := by simp [Measure.dirac_apply' mean (MeasurableSet.singleton (x i₁)), hne] exact ENNReal.ofReal_ne_zero_iff.mpr (hp i₁) (hmass₁.symm.trans hzero) exact hx₀₁ (hmean₀.symm.trans hmean₁) · have hnoatoms : NullSingletonClass (atomicLaw n x p) := by rw [hv] exact nullSingletonClass_gaussianReal hvzero have hzero : atomicLaw n x p {x i₀} = 0 := hnoatoms.measure_singleton (x i₀) rw [atomicLaw_singleton hx hpnonneg i₀] at hzero exact ENNReal.ofReal_ne_zero_iff.mpr (hp i₀) hzero
Causalean.Stat.MomentProblems.not_isGaussianLaw_atomicLaw · Causalean/Stat/Nonparametric/MomentProblems/AtomicLaw.lean:115 · uses IsGaussianLaw , atomicLaw
4 supporting declarations (lemmas, instances)
Symmetric­Atom­Solve 11 core · 15 supporting Fix L + 1 equally spaced atoms placed symmetrically about the origin. ★ continuous_cumulantToWeights★ cumulantToWeights_uniformCumulantPoint

Solving for atom weights that match a prescribed cumulant target

Fix L + 1 equally spaced atoms placed symmetrically about the origin. Matching the raw moments of orders 0, 1, …, L of a weight vector supported on those atoms to a prescribed moment target is a square linear system whose matrix is a transposed Vandermonde matrix in the atom locations. The atoms being distinct, that matrix is invertible, so the weights are recovered from the moment target by a linear — in particular continuous — solve.

Composing this solve with the triangular moment↔cumulant inversion gives the map this module exports: prescribed truncated cumulants ↦ atom weights, which is continuous, and which sends the cumulants of the uniform-weight law back to the uniform weights. That is exactly what is needed to conclude, by openness of the strict-positivity constraint, that a whole neighborhood of cumulant targets is matched by genuine (positive-weight) probability laws.

Nothing here yet asserts positivity of the solved weights: it is a pure linear-algebra and continuity layer. The realizability conclusion is drawn in TruncatedCumulantInterior.

def symmetricAtoms reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L, the symmetric atom locations are the L + 1 real numbers −L, −L + 2, …, L, indexed in increasing order; equivalently, the location with index i is twice i minus L.

Definition (Lean source)
L :
symmetricAtoms L :
Fin (L + 1) → ℝ
fun i => 2 * (i.val : ℝ) - (L : ℝ)
Causalean.Stat.MomentProblems.symmetricAtoms · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:34
def uniformWeights reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L, the uniform weight vector assigns mass 1/(L+1)1/(L+1) to each of the L + 1 symmetric atoms.

Definition (Lean source)
L :
uniformWeights L :
Fin (L + 1) → ℝ
fun _ => 1 / (L + 1 : ℝ)
Causalean.Stat.MomentProblems.uniformWeights · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:68
def uniformMoments reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L, the raw-moment sequence of the uniform law on the symmetric atoms assigns to every nonnegative order the equally weighted average of that power of the L + 1 symmetric atom locations.

Definition (Lean source)
L :
uniformMoments L :
ℕ → ℝ
fun k => ∑ i, uniformWeights L i * symmetricAtoms L i ^ k
Causalean.Stat.MomentProblems.uniformMoments · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:91
def uniformCumulants reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L, the cumulant sequence of the uniform law on the symmetric atoms is the cumulant sequence calculated from that law's raw moments.

Definition (Lean source)
L :
uniformCumulants L :
ℕ → ℝ
fun r => cumFromMom r (uniformMoments L)
Causalean.Stat.MomentProblems.uniformCumulants · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:105
def atomVandermonde reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L, the Vandermonde matrix of the symmetric atoms is the square real matrix whose row for each atom lists successive powers of that atom, from power zero through power L.

Definition (Lean source)
L :
atomVandermonde L :
Matrix (Fin (L + 1)) (Fin (L + 1)) ℝ
Matrix.vandermonde (symmetricAtoms L)
Causalean.Stat.MomentProblems.atomVandermonde · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:122
def atomSolve reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L and a prescribed vector of real raw moments of orders zero through L, the moment-matching weight vector is the result of applying the inverse transpose of the Vandermonde matrix of the symmetric atoms to that vector.

Definition (Lean source)
L :
b :
Fin (L + 1) → ℝ
atomSolve L b :
Fin (L + 1) → ℝ
(atomVandermonde L).transpose⁻¹.mulVec b
def padCumulants reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L and a prescribed vector of cumulants of orders zero through L, the zero-padded cumulant sequence agrees with that vector through order L and equals zero at every higher order.

Definition (Lean source)
L :
y :
Fin (L + 1) → ℝ
padCumulants L y :
ℕ → ℝ
fun k => if h : k < L + 1 then y ⟨k, h⟩ else 0
def cumulantToWeights reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L and a prescribed vector of cumulants of orders zero through L, the cumulant-to-weights map first reconstructs the corresponding raw moments and then returns the symmetric-atom weights obtained by solving their Vandermonde moment-matching system.

Definition (Lean source)
L :
y :
Fin (L + 1) → ℝ
cumulantToWeights L y :
Fin (L + 1) → ℝ
atomSolve L (fun k => momFromCum (padCumulants L y) k.val)
Causalean.Stat.MomentProblems.cumulantToWeights · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:209
theorem continuous_cumulantToWeights reviewed
Causalean.Stat.MomentProblems

The cumulant-to-weights map is continuous. For any truncation order L, the map sending a truncated cumulant target to the recovered symmetric-atom weights is continuous, since both the moment↔cumulant inversion and the Vandermonde solve it composes are continuous.

Formal statement
L :
Proof (Lean source)
@[fun_prop] theorem continuous_cumulantToWeights (L : ℕ) : Continuous (cumulantToWeights L) := by apply (continuous_atomSolve L).comp apply continuous_pi intro k exact (continuous_momFromCum k.val).comp (continuous_padCumulants L)
Causalean.Stat.MomentProblems.continuous_cumulantToWeights · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:217 · uses cumulantToWeights
def uniformCumulantPoint reviewed
Causalean.Stat.MomentProblems

For a nonnegative integer L, the base cumulant target is the vector of cumulants of orders zero through L of the uniform probability law on the symmetric atoms.

Definition (Lean source)
L :
uniformCumulantPoint L :
Fin (L + 1) → ℝ
fun k => uniformCumulants L k.val
Causalean.Stat.MomentProblems.uniformCumulantPoint · Causalean/Stat/Nonparametric/MomentProblems/SymmetricAtomSolve.lean:227
theorem cumulantToWeights_uniformCumulantPoint reviewed
Causalean.Stat.MomentProblems

The base point maps to the uniform weights. For any truncation order L, feeding the cumulants of the uniform-weight law on the symmetric atoms into the cumulant-to-weights map returns exactly the uniform weights — all strictly positive, hence strictly inside the positivity constraints.

Formal statement
L :
cumulantToWeights L (uniformCumulantPoint L) = uniformWeights L
Proof (Lean source)
theorem cumulantToWeights_uniformCumulantPoint (L : ℕ) : cumulantToWeights L (uniformCumulantPoint L) = uniformWeights L := by calc cumulantToWeights L (uniformCumulantPoint L) = atomSolve L (fun k => momFromCum (padCumulants L (uniformCumulantPoint L)) k.val) := rfl _ = atomSolve L (fun k => uniformMoments L k.val) := by congr 1 funext k exact momFromCum_padCumulants_uniform L k _ = uniformWeights L := atomSolve_uniformMoments L
15 supporting declarations (lemmas, instances)
Truncated­Cumulant­Interior 2 core · 0 supporting The main result of this module says that the truncated cumulant sequences of orders two through L attainable by *centered, non-Gaussian* probability laws on the real line with finite L-th moment fill up a set with nonemp ★ truncatedMomentInterior

The truncated cumulant range has nonempty interior

The main result of this module says that the truncated cumulant sequences of orders two through L attainable by centered, non-Gaussian probability laws on the real line with finite L-th moment fill up a set with nonempty interior: there is one attainable cumulant vector all of whose sufficiently small perturbations are again attainable by such a law.

The proof is constructive rather than an appeal to the general (truncated Hamburger) moment problem. Put L + 1 distinct equally spaced symmetric atoms on the line. Matching the first L + 1 raw moments of a weight vector to a target is then a linear system with a transposed Vandermonde matrix, hence invertible, and the moment target itself is obtained from the cumulant target by the triangular moment↔cumulant inversion — so weights depend continuously on the prescribed cumulants (SymmetricAtomSolve). At the uniform weight vector all weights are strictly positive, and positivity is an open condition, so every nearby cumulant target still solves to strictly positive weights, i.e. to a genuine probability law. That law is centered (the atoms are symmetric and the first-moment equation is part of the solved system), has all moments finite (finite support), and is non-Gaussian (at least two atoms carry positive mass).

The order L = 0 case is degenerate — a single atom would be a point mass, i.e. a degenerate Gaussian — and is handled separately by an explicit fair two-point law on ±1.

References: Akhiezer (1965), The Classical Moment Problem; Curto–Fialkow (1991) on truncated moment problems and flat extensions; Schmüdgen (2017), The Moment Problem, Theorem 10.7.

def TruncatedMomentInterior reviewed
Causalean.Stat.MomentProblems

For a nonnegative truncation order, the interior-of-the-truncated-cumulant-range condition says that there exist a real cumulant sequence and a strictly positive real radius such that every real sequence whose entries of every order from two through the truncation order differ from the corresponding target entries by less than that radius is the cumulant sequence, at each of those orders, of some centered non-Gaussian probability law on the real line with finite moment of the truncation order.

Definition (Lean source)
L :
TruncatedMomentInterior L :
Prop
∃ (c : ℕ → ℝ) (ε : ℝ),
0 < ε
clause 1
c' :
ℕ → ℝ
(∀ r, 2 ≤ r → r ≤ L → |c' r - c r| < ε)
∃ ν : Measure ℝ,
∫ x, x ∂ν = 0
MemLp (id : ℝ → ℝ) (L : ℝ≥0∞) ν
r :
2 ≤ r
r ≤ L
sourceCumulant ν id r = c' r
Causalean.Stat.MomentProblems.TruncatedMomentInterior · Causalean/Stat/Nonparametric/MomentProblems/TruncatedCumulantInterior.lean:40
theorem truncatedMomentInterior reviewed
Causalean.Stat.MomentProblems

The truncated cumulant range has nonempty interior. For every truncation order L, there is a cumulant vector, of orders two through L, together with a strictly positive radius, such that every cumulant vector within that radius is the truncated cumulant vector of some centered, non-Gaussian probability law on the real line with finite L-th moment.

Formal statement
L :
Proof (Lean source)
theorem truncatedMomentInterior (L : ℕ) : TruncatedMomentInterior L := by by_cases hL : L = 0 · simpa [hL] using truncatedMomentInterior_zero have hLpos : 1 ≤ L := by omega let radius : ℝ := 1 / (L + 1 : ℝ) have hradius : 0 < radius := by dsimp [radius]; positivity obtain ⟨ε, hε, hcont⟩ := Metric.continuousAt_iff.mp (continuous_cumulantToWeights L).continuousAt radius hradius refine ⟨uniformCumulants L, ε, hε, ?_⟩ intro c' hc' let y : Fin (L + 1) → ℝ := fun k => if 2 ≤ k.val then c' k.val else uniformCumulants L k.val have hy : dist y (uniformCumulantPoint L) < ε := by rw [dist_pi_lt_iff hε] intro k by_cases hk : 2 ≤ k.val · simpa [y, uniformCumulantPoint, hk, Real.dist_eq] using hc' k.val hk (by omega) · simpa [y, uniformCumulantPoint, hk] using hε have hphiDist : dist (cumulantToWeights L y) (cumulantToWeights L (uniformCumulantPoint L)) < radius := hcont hy have hp : ∀ i, 0 < cumulantToWeights L y i := by intro i have hi := (dist_pi_lt_iff hradius).mp hphiDist i rw [cumulantToWeights_uniformCumulantPoint L] at hi simp only [Real.dist_eq, uniformWeights, radius] at hi linarith [abs_lt.mp hi] have hpad : ∀ k, 2 ≤ k → k ≤ L → padCumulants L y k = c' k := by intro k hk hkL simp [padCumulants, y, hk, show k < L + 1 by omega] have hmom : ∀ k, k ≤ L → momFromCum (padCumulants L y) k = momFromCum c' k := by intro k hk apply momFromCum_congr intro j hj hjk exact hpad j hj (hjk.trans hk) have hsolve : ∀ k, k ≤ L → ∑ i, cumulantToWeights L y i * symmetricAtoms L i ^ k = momFromCum c' k := by intro k hk let kf : Fin (L + 1) := ⟨k, by omega⟩ simpa [cumulantToWeights, kf, hmom k hk] using atomSolve_spec L (fun j => momFromCum (padCumulants L y) j.val) kf have hsum : ∑ i, cumulantToWeights L y i = 1 := by simpa using hsolve 0 (by omega) have hmean : ∑ i, cumulantToWeights L y i * symmetricAtoms L i = 0 := by simpa using hsolve 1 hLpos let ν := atomicLaw (L + 1) (symmetricAtoms L) (cumulantToWeights L y) have hpnonneg : ∀ i, 0 ≤ cumulantToWeights L y i := fun i => (hp i).le refine ⟨ν, isProbabilityMeasure_atomicLaw (symmetricAtoms_injective L) hpnonneg hsum, ?_, not_isGaussianLaw_atomicLaw (symmetricAtoms_injective L) hp (by omega), memLp_id_atomicLaw (symmetricAtoms_injective L) hpnonneg hsum (L : ℝ≥0∞), ?_⟩ · simpa [ν] using (integral_atomicLaw (symmetricAtoms_injective L) hpnonneg id).trans hmean · intro r hr hrL rw [sourceCumulant_eq_cumFromMom] calc cumFromMom r (fun k => ∫ t, t ^ k ∂ν) = cumFromMom r (momFromCum c') := by apply cumFromMom_congr intro k hk hkR rw [integral_atomicLaw (symmetricAtoms_injective L) hpnonneg] exact hsolve k (hkR.trans hrL) _ = c' r := cumFromMom_momFromCum c' r hr