redistribution_characteristic_imaginary_log_bound
For any real reports and any real y, the characteristic value at i*y is strictly positive in norm and its negative logarithm is at most y times the empirical mean.
Statement
theorem redistribution_characteristic_imaginary_log_bound {n : ℕ} (hn : 0 < n)
(θ : Fin n → ℝ) (y : ℝ) :
0 < ‖redistribution_characteristic θ ((y : ℂ)*Complex.I)‖ ∧
-Real.log ‖redistribution_characteristic θ ((y : ℂ)*Complex.I)‖ ≤
y * (∑ j, θ j) / (n : ℝ)- Source
- Routine finite-characteristic-function helper for IN-251, arXiv:2610.10995 Lemma4.3 Steps1/2 and equation12. These statements do not assert the half-plane boundary integral inequality or the main harmonic error bound. No research advance.
- Verified
- 27 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 1
- redistribution_characteristicThe empirical characteristic function continued to the complex plane: the normalized finite sum of exp(i*z*θ_j).
The proof, explained
Evaluating the redistribution characteristic at the imaginary argument i * y reduces the complex exponentials to real exponentials, yielding a positive real average S = (1 / n) * ∑ j, exp(-y * θ j) whose norm is S. The inequality -log(S) ≤ y * (∑ j, θ j) / n is Jensen's inequality for the convex function x ↦ exp(x), proved directly by bounding each term from below using the tangent line at m = -(y * ∑ j, θ j) / n. Summing these linear bounds across all n coordinates cancels the deviations from the mean, proving exp(m) ≤ S, which upon taking logarithms and negating produces the bound.
The characteristic function evaluated at i * y equals the real average S = (1 / n) * ∑ j, exp(-y * θ j).
Using i * (i * y) * θ j = i² * y * θ j = -y * θ j, each complex exponential exp(i * (i * y) * θ j) reduces to the real exponential exp(-y * θ j). Unfolding the definition of redistribution_characteristic and summing these terms shows that the value at i * y is exactly the real number S cast to ℂ.
uses
Nat.cast_pos.mpr,Finset.sum_congr,Complex.I_mul_Ihave hn_pos : 0 < (n : ℝ) := Nat.cast_pos.mpr hn have hn_ne : (n : ℝ) ≠ 0 := hn_pos.ne' set S := 1 / (n : ℝ) * ∑ j, Real.exp (-y * θ j) with hS_def have h_char : redistribution_characteristic θ ((y : ℂ) * Complex.I) = (S : ℂ) := by rw [hS_def] unfold redistribution_characteristic push_cast congr 1 apply Finset.sum_congr rfl intro j _ have h1 : Complex.I * ((y : ℂ) * Complex.I) * (θ j : ℂ) = - (y : ℂ) * (θ j : ℂ) := by calc Complex.I * ((y : ℂ) * Complex.I) * (θ j : ℂ) = (Complex.I * Complex.I) * (y : ℂ) * (θ j : ℂ) := by ring _ = -1 * (y : ℂ) * (θ j : ℂ) := by rw [Complex.I_mul_I] _ = -(y : ℂ) * (θ j : ℂ) := by ring rw [h1]
The average S is strictly positive, its norm is S, and the first conjunct 0 < ‖characteristic‖ holds.
Because n > 0, the index set Fin n is nonempty. Since the exponential of any real number is strictly positive, the sum of exponentials is strictly positive, and multiplying by 1 / n > 0 gives S > 0. The complex norm of S is S itself, establishing 0 < ‖characteristic‖ and reducing the remaining goal to -log(S) ≤ y * (∑ j, θ j) / n.
uses
Finset.univ_nonempty_iff.mpr,Fin.pos_iff_nonempty.mp,Finset.sum_pos,Real.exp_pos,mul_pos,one_div_pos.mpr,Complex.norm_real,Real.norm_of_nonneghave h_univ : (Finset.univ : Finset (Fin n)).Nonempty := Finset.univ_nonempty_iff.mpr (Fin.pos_iff_nonempty.mp hn) have h_sum_pos : 0 < ∑ j, Real.exp (-y * θ j) := Finset.sum_pos (fun j _ => Real.exp_pos (-y * θ j)) h_univ have hS_pos : 0 < S := mul_pos (one_div_pos.mpr hn_pos) h_sum_pos have h_norm : ‖redistribution_characteristic θ ((y : ℂ) * Complex.I)‖ = S := by rw [h_char, Complex.norm_real, Real.norm_of_nonneg hS_pos.le] refine ⟨by rw [h_norm]; exact hS_pos, ?_⟩ rw [h_norm]
For each index j, the tangent line lower bound exp(m) * (-y * θ j - m + 1) ≤ exp(-y * θ j) holds, where m = -(y * ∑ j, θ j) / n.
From the convexity inequality u + 1 ≤ exp(u) applied to u = -y * θ j - m, we multiply both sides by the positive constant exp(m). Simplifying exp(m) * exp(-y * θ j - m) = exp(-y * θ j) yields the tangent line bound at m for each report θ j.
uses
mul_le_mul_of_nonneg_left,Real.add_one_le_exp,Real.exp_pos,Real.exp_addset m := - (y * (∑ j, θ j) / (n : ℝ)) with hm_def have h_exp_term : ∀ j : Fin n, Real.exp m * (-y * θ j - m + 1) ≤ Real.exp (-y * θ j) := by intro j have h1 := mul_le_mul_of_nonneg_left (Real.add_one_le_exp (-y * θ j - m)) (Real.exp_pos m).le have h2 : Real.exp m * Real.exp (-y * θ j - m) = Real.exp (-y * θ j) := by rw [← Real.exp_add] congr 1 ring rw [h2] at h1 exact h1
Summing the pointwise tangent bounds yields exp(m) * ∑ j, (-y * θ j - m + 1) ≤ ∑ j, exp(-y * θ j).
Summing the pointwise inequality over all j in Fin n and factoring the common factor exp(m) out of the sum preserves the inequality.
uses
Finset.sum_le_sum,Finset.mul_sumhave h_sum : ∑ j : Fin n, (Real.exp m * (-y * θ j - m + 1)) ≤ ∑ j : Fin n, Real.exp (-y * θ j) := Finset.sum_le_sum (fun j _ => h_exp_term j) rw [← Finset.mul_sum] at h_sum
The inner sum ∑ j, (-y * θ j - m + 1) evaluates to n, so exp(m) * n ≤ ∑ j, exp(-y * θ j).
Splitting the sum into ∑ j, (-y * θ j) and ∑ j, (1 - m), the first part is -y * ∑ j, θ j and the second is n * (1 - m) = n - n * m. Since m = -(y * ∑ j, θ j) / n, the product n * m cancels with -y * ∑ j, θ j, leaving exactly n.
uses
Finset.sum_add_distrib,Finset.sum_congr,Finset.mul_sum,Finset.sum_const,Finset.card_univ,Fintype.card_fin,nsmul_eq_mulhave h_inner : ∑ j : Fin n, (-y * θ j - m + 1) = (n : ℝ) := by have h1 : ∑ j : Fin n, (-y * θ j - m + 1) = (∑ j : Fin n, -y * θ j) + ∑ j : Fin n, (1 - m) := by rw [← Finset.sum_add_distrib] apply Finset.sum_congr rfl intro j _ ring rw [h1] rw [← Finset.mul_sum] rw [Finset.sum_const, Finset.card_univ, Fintype.card_fin] rw [nsmul_eq_mul] rw [hm_def] field_simp ring rw [h_inner] at h_sum
exp(m) ≤ S.
Dividing both sides of exp(m) * n ≤ ∑ j, exp(-y * θ j) by n (or multiplying by 1 / n ≥ 0) yields exp(m) ≤ (1 / n) * ∑ j, exp(-y * θ j) = S.
uses
mul_assoc,mul_one_div_cancel,mul_one,mul_le_mul_of_nonneg_right,one_div_nonneg.mprhave h_exp_le : Real.exp m ≤ S := by rw [hS_def] calc Real.exp m = Real.exp m * (n : ℝ) * (1 / (n : ℝ)) := by rw [mul_assoc, mul_one_div_cancel hn_ne, mul_one] _ ≤ (∑ j : Fin n, Real.exp (-y * θ j)) * (1 / (n : ℝ)) := mul_le_mul_of_nonneg_right h_sum (one_div_nonneg.mpr hn_pos.le) _ = 1 / (n : ℝ) * ∑ j : Fin n, Real.exp (-y * θ j) := by ring
-log(S) ≤ y * (∑ j, θ j) / n, which completes the proof.
Since S > 0 and exp(m) ≤ S, taking logarithms gives m ≤ log(S). Negating both sides reverses the inequality to -log(S) ≤ -m. Substituting m = -(y * ∑ j, θ j) / n and cancelling the double negation gives the desired upper bound.
uses
Real.le_log_iff_exp_le,neg_le_neg,neg_neghave hm_le_log : m ≤ Real.log S := (Real.le_log_iff_exp_le hS_pos).mpr h_exp_le have h_neg : -Real.log S ≤ -m := neg_le_neg hm_le_log rw [hm_def] at h_neg have h_cancel : -(- (y * (∑ j, θ j) / (n : ℝ))) = y * (∑ j, θ j) / (n : ℝ) := neg_neg _ rw [h_cancel] at h_neg exact h_neg
Complex.I_mul_I- The imaginary unit satisfies I * I = -1.
Complex.norm_real- The complex norm of a real number embedded in ℂ equals its absolute value in ℝ.
Fin.pos_iff_nonempty.mp- The finite type Fin n is nonempty if 0 < n.
Finset.card_univ- The cardinality of Finset.univ for a finite type is the cardinality of the type.
Finset.mul_sum- Multiplication distributes over a finite sum: a * ∑ i, f i = ∑ i, a * f i.
Finset.sum_add_distrib- The sum of a sum over a finite set is the sum of the individual sums: ∑ i, (f i + g i) = ∑ i, f i + ∑ i, g i.
Finset.sum_congr- If two finite index sets are equal and the summands agree on all elements, their sums are equal.
Finset.sum_const- The sum of a constant c over a finite set s equals |s| • c.
Finset.sum_le_sum- If f i ≤ g i for all i in a finite set, then ∑ i, f i ≤ ∑ i, g i.
Finset.sum_pos- A sum of strictly positive elements over a nonempty finite set is strictly positive.
Finset.univ_nonempty_iff.mpr- The universal finset Finset.univ is nonempty if the underlying type is nonempty.
Fintype.card_fin- The cardinality of Fin n is n.
Nat.cast_pos.mpr- A natural number is strictly positive in ℝ if it is strictly positive in ℕ.
Real.exp_pos- The exponential of any real number is strictly positive: exp(x) > 0.
Real.exp_add- The exponential of a sum is the product of exponentials: exp(x + y) = exp(x) * exp(y).
Real.add_one_le_exp- For any real number x, x + 1 ≤ exp(x).
Real.le_log_iff_exp_le- For y > 0, x ≤ log(y) if and only if exp(x) ≤ y.
Real.norm_of_nonneg- For a non-negative real number r ≥ 0, ‖r‖ = r.
mul_assoc- Multiplication is associative: (a * b) * c = a * (b * c).
mul_le_mul_of_nonneg_left- If b ≤ c and 0 ≤ a, then a * b ≤ a * c.
mul_le_mul_of_nonneg_right- If b ≤ c and 0 ≤ a, then b * a ≤ c * a.
mul_one- Multiplying an element by 1 leaves it unchanged: a * 1 = a.
mul_one_div_cancel- For any non-zero element a ≠ 0, a * (1 / a) = 1.
mul_pos- The product of two strictly positive elements is strictly positive.
neg_le_neg- If a ≤ b, then -b ≤ -a.
neg_neg- Double negation cancels: -(-a) = a.
nsmul_eq_mul- Scaling by a natural number n • a in ℝ is equivalent to multiplying by (n : ℝ) * a.
one_div_nonneg.mpr- If 0 ≤ x, then 0 ≤ 1 / x.
one_div_pos.mpr- If 0 < x, then 0 < 1 / x.
Written by a model (gemini-3.8-flash) from the Lean proof and the goal states Lean reported at each step. Lean verified the proof; this explanation is a reading aid and is not itself machine-checked.
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
open Finset MeasureTheorytheorem redistribution_characteristic_imaginary_log_bound {n : ℕ} (hn : 0 < n) (θ : Fin n → ℝ) (y : ℝ) : 0 < ‖redistribution_characteristic θ ((y : ℂ)*Complex.I)‖ ∧ -Real.log ‖redistribution_characteristic θ ((y : ℂ)*Complex.I)‖ ≤ y * (∑ j, θ j) / (n : ℝ) := by have hn_pos : 0 < (n : ℝ) := Nat.cast_pos.mpr hn have hn_ne : (n : ℝ) ≠ 0 := hn_pos.ne' set S := 1 / (n : ℝ) * ∑ j, Real.exp (-y * θ j) with hS_def have h_char : redistribution_characteristic θ ((y : ℂ) * Complex.I) = (S : ℂ) := by rw [hS_def] unfold redistribution_characteristic push_cast congr 1 apply Finset.sum_congr rfl intro j _ have h1 : Complex.I * ((y : ℂ) * Complex.I) * (θ j : ℂ) = - (y : ℂ) * (θ j : ℂ) := by calc Complex.I * ((y : ℂ) * Complex.I) * (θ j : ℂ) = (Complex.I * Complex.I) * (y : ℂ) * (θ j : ℂ) := by ring _ = -1 * (y : ℂ) * (θ j : ℂ) := by rw [Complex.I_mul_I] _ = -(y : ℂ) * (θ j : ℂ) := by ring rw [h1] have h_univ : (Finset.univ : Finset (Fin n)).Nonempty := Finset.univ_nonempty_iff.mpr (Fin.pos_iff_nonempty.mp hn) have h_sum_pos : 0 < ∑ j, Real.exp (-y * θ j) := Finset.sum_pos (fun j _ => Real.exp_pos (-y * θ j)) h_univ have hS_pos : 0 < S := mul_pos (one_div_pos.mpr hn_pos) h_sum_pos have h_norm : ‖redistribution_characteristic θ ((y : ℂ) * Complex.I)‖ = S := by rw [h_char, Complex.norm_real, Real.norm_of_nonneg hS_pos.le] refine ⟨by rw [h_norm]; exact hS_pos, ?_⟩ rw [h_norm] set m := - (y * (∑ j, θ j) / (n : ℝ)) with hm_def have h_exp_term : ∀ j : Fin n, Real.exp m * (-y * θ j - m + 1) ≤ Real.exp (-y * θ j) := by intro j have h1 := mul_le_mul_of_nonneg_left (Real.add_one_le_exp (-y * θ j - m)) (Real.exp_pos m).le have h2 : Real.exp m * Real.exp (-y * θ j - m) = Real.exp (-y * θ j) := by rw [← Real.exp_add] congr 1 ring rw [h2] at h1 exact h1 have h_sum : ∑ j : Fin n, (Real.exp m * (-y * θ j - m + 1)) ≤ ∑ j : Fin n, Real.exp (-y * θ j) := Finset.sum_le_sum (fun j _ => h_exp_term j) rw [← Finset.mul_sum] at h_sum have h_inner : ∑ j : Fin n, (-y * θ j - m + 1) = (n : ℝ) := by have h1 : ∑ j : Fin n, (-y * θ j - m + 1) = (∑ j : Fin n, -y * θ j) + ∑ j : Fin n, (1 - m) := by rw [← Finset.sum_add_distrib] apply Finset.sum_congr rfl intro j _ ring rw [h1] rw [← Finset.mul_sum] rw [Finset.sum_const, Finset.card_univ, Fintype.card_fin] rw [nsmul_eq_mul] rw [hm_def] field_simp ring rw [h_inner] at h_sum have h_exp_le : Real.exp m ≤ S := by rw [hS_def] calc Real.exp m = Real.exp m * (n : ℝ) * (1 / (n : ℝ)) := by rw [mul_assoc, mul_one_div_cancel hn_ne, mul_one] _ ≤ (∑ j : Fin n, Real.exp (-y * θ j)) * (1 / (n : ℝ)) := mul_le_mul_of_nonneg_right h_sum (one_div_nonneg.mpr hn_pos.le) _ = 1 / (n : ℝ) * ∑ j : Fin n, Real.exp (-y * θ j) := by ring have hm_le_log : m ≤ Real.log S := (Real.le_log_iff_exp_le hS_pos).mpr h_exp_le have h_neg : -Real.log S ≤ -m := neg_le_neg hm_le_log rw [hm_def] at h_neg have h_cancel : -(- (y * (∑ j, θ j) / (n : ℝ))) = y * (∑ j, θ j) / (n : ℝ) := neg_neg _ rw [h_cancel] at h_neg exact h_neg