9.3. Single-bit hypercontractivity
-
FABL.uniformLInfinityNorm[complete] -
FABL.uniformLpNormENNReal[complete] -
FABL.uniformLpNormENNReal_coe[complete] -
FABL.IsHypercontractiveRandomVariable[complete]
Definition 9.13. Let 1\le p\le q\le\infty and 0\le\rho<1.
A real random variable X with finite q-norm is
(p,q,\rho)-hypercontractive when, for all a,b\in\mathbb R,
\|a+\rho bX\|_q\le\|a+bX\|_p.
Lean code for Definition9.3.1●4 declarations
Associated Lean declarations
-
FABL.uniformLInfinityNorm[complete]
-
FABL.uniformLpNormENNReal[complete]
-
FABL.uniformLpNormENNReal_coe[complete]
-
FABL.IsHypercontractiveRandomVariable[complete]
-
FABL.uniformLInfinityNorm[complete] -
FABL.uniformLpNormENNReal[complete] -
FABL.uniformLpNormENNReal_coe[complete] -
FABL.IsHypercontractiveRandomVariable[complete]
-
defdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
def FABL.uniformLInfinityNorm.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (f : Ω → ℝ) : ℝ
def FABL.uniformLInfinityNorm.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (f : Ω → ℝ) : ℝ
The actual uniform `L∞` norm on a nonempty finite domain.
-
defdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
def FABL.uniformLpNormENNReal.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q : ENNReal) (f : Ω → ℝ) : ℝ
def FABL.uniformLpNormENNReal.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q : ENNReal) (f : Ω → ℝ) : ℝ
A finite uniform exponent norm which agrees with the established real-exponent API at every finite exponent and uses the maximum at `∞`.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformLpNormENNReal_coe.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q : NNReal) (f : Ω → ℝ) : FABL.uniformLpNormENNReal (↑q) f = FABL.uniformLpNorm (↑q) f
theorem FABL.uniformLpNormENNReal_coe.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q : NNReal) (f : Ω → ℝ) : FABL.uniformLpNormENNReal (↑q) f = FABL.uniformLpNorm (↑q) f
Compatibility with FABL's existing `uniformLpNorm` at a finite `NNReal` exponent.
-
defdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
def FABL.IsHypercontractiveRandomVariable.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (p q : ENNReal) (ρ : ℝ) (X : Ω → ℝ) : Prop
def FABL.IsHypercontractiveRandomVariable.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (p q : ENNReal) (ρ : ℝ) (X : Ω → ℝ) : Prop
O'Donnell, Definition 9.13, on a finite probability space.
-
FABL.uniformLpNorm_const_mul[complete] -
FABL.uniformLInfinityNorm_const_mul[complete] -
FABL.uniformLpNormENNReal_const_mul[complete] -
FABL.IsHypercontractiveRandomVariable.of_normalized[complete]
Remark 9.14. By homogeneity it is enough to check the defining inequality
with a=1 or with b=1. If it holds at \rho, it also holds at every
0\le\rho'\le\rho.
Lean code for Lemma9.3.2●4 theorems
Associated Lean declarations
-
FABL.uniformLpNorm_const_mul[complete]
-
FABL.uniformLInfinityNorm_const_mul[complete]
-
FABL.uniformLpNormENNReal_const_mul[complete]
-
FABL.IsHypercontractiveRandomVariable.of_normalized[complete]
-
FABL.uniformLpNorm_const_mul[complete] -
FABL.uniformLInfinityNorm_const_mul[complete] -
FABL.uniformLpNormENNReal_const_mul[complete] -
FABL.IsHypercontractiveRandomVariable.of_normalized[complete]
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformLpNorm_const_mul.{u_1} {Ω : Type u_1} [Fintype Ω] (p c : ℝ) (f : Ω → ℝ) (hp : 0 < p) : (FABL.uniformLpNorm p fun x => c * f x) = |c| * FABL.uniformLpNorm p f
theorem FABL.uniformLpNorm_const_mul.{u_1} {Ω : Type u_1} [Fintype Ω] (p c : ℝ) (f : Ω → ℝ) (hp : 0 < p) : (FABL.uniformLpNorm p fun x => c * f x) = |c| * FABL.uniformLpNorm p f
Positive finite-exponent uniform norms are absolutely homogeneous.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformLInfinityNorm_const_mul.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (c : ℝ) (f : Ω → ℝ) : (FABL.uniformLInfinityNorm fun x => c * f x) = |c| * FABL.uniformLInfinityNorm f
theorem FABL.uniformLInfinityNorm_const_mul.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (c : ℝ) (f : Ω → ℝ) : (FABL.uniformLInfinityNorm fun x => c * f x) = |c| * FABL.uniformLInfinityNorm f
The actual finite-uniform `L∞` norm is absolutely homogeneous.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformLpNormENNReal_const_mul.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q : ENNReal) (c : ℝ) (f : Ω → ℝ) (hq : q ≠ 0) : (FABL.uniformLpNormENNReal q fun x => c * f x) = |c| * FABL.uniformLpNormENNReal q f
theorem FABL.uniformLpNormENNReal_const_mul.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q : ENNReal) (c : ℝ) (f : Ω → ℝ) (hq : q ≠ 0) : (FABL.uniformLpNormENNReal q fun x => c * f x) = |c| * FABL.uniformLpNormENNReal q f
The `ENNReal` finite-uniform interface is absolutely homogeneous at every nonzero exponent, including `∞`.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.IsHypercontractiveRandomVariable.of_normalized.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] {p q : ENNReal} (hp : p ≠ 0) (hq : q ≠ 0) {ρ : ℝ} {X : Ω → ℝ} (h : ∀ (a b : ℝ), a = 1 ∨ b = 1 → (FABL.uniformLpNormENNReal q fun x => a + ρ * b * X x) ≤ FABL.uniformLpNormENNReal p fun x => a + b * X x) : FABL.IsHypercontractiveRandomVariable p q ρ X
theorem FABL.IsHypercontractiveRandomVariable.of_normalized.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] {p q : ENNReal} (hp : p ≠ 0) (hq : q ≠ 0) {ρ : ℝ} {X : Ω → ℝ} (h : ∀ (a b : ℝ), a = 1 ∨ b = 1 → (FABL.uniformLpNormENNReal q fun x => a + ρ * b * X x) ≤ FABL.uniformLpNormENNReal p fun x => a + b * X x) : FABL.IsHypercontractiveRandomVariable p q ρ X
Remark 9.14, homogeneity reduction: it is enough to check affine pairs for which one coefficient is one.
Proposition 9.15. If independent random variables X and Y are both
(p,q,\rho)-hypercontractive, then X+Y is
(p,q,\rho)-hypercontractive. The book defers the proof to
Exercise 10.2; this node supplies no assumption.
-
FABL.oneBit_twoSix[complete] -
FABL.oneBit_twoSix_necessary_sq[complete] -
FABL.oneBit_twoSix_iff[complete] -
FABL.uniformSign_twoSixHypercontractive[complete]
Proposition 9.16. For a uniform sign x,
\|a+\rho bx\|_6\le\|a+bx\|_2
\quad\text{for all }a,b\in\mathbb R
if and only if
\rho\le\frac1{\sqrt5}.
Lean code for Theorem9.3.4●4 theorems
Associated Lean declarations
-
FABL.oneBit_twoSix[complete]
-
FABL.oneBit_twoSix_necessary_sq[complete]
-
FABL.oneBit_twoSix_iff[complete]
-
FABL.uniformSign_twoSixHypercontractive[complete]
-
FABL.oneBit_twoSix[complete] -
FABL.oneBit_twoSix_necessary_sq[complete] -
FABL.oneBit_twoSix_iff[complete] -
FABL.uniformSign_twoSixHypercontractive[complete]
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_twoSix (a b ρ : ℝ) (hρ : ρ ^ 2 ≤ 1 / 5) : (FABL.uniformLpNorm 6 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
theorem FABL.oneBit_twoSix (a b ρ : ℝ) (hρ : ρ ^ 2 ≤ 1 / 5) : (FABL.uniformLpNorm 6 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
O'Donnell, Proposition 9.16, sufficient direction with the full sharp interval `0 ≤ ρ ≤ 1/√5`.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_twoSix_necessary_sq (ρ : ℝ) (h : ∀ (a b : ℝ), (FABL.uniformLpNorm 6 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x) : ρ ^ 2 ≤ 1 / 5
theorem FABL.oneBit_twoSix_necessary_sq (ρ : ℝ) (h : ∀ (a b : ℝ), (FABL.uniformLpNorm 6 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x) : ρ ^ 2 ≤ 1 / 5
The quadratic term in Proposition 9.16 is necessary. This is the small-`b` argument from the book, expressed as a limit along `b = 1/(m+1)`.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_twoSix_iff (ρ : ℝ) (hρ : 0 ≤ ρ) : (∀ (a b : ℝ), (FABL.uniformLpNorm 6 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x) ↔ ρ ≤ FABL.twoSixNoiseRoot
theorem FABL.oneBit_twoSix_iff (ρ : ℝ) (hρ : 0 ≤ ρ) : (∀ (a b : ℝ), (FABL.uniformLpNorm 6 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x) ↔ ρ ≤ FABL.twoSixNoiseRoot
O'Donnell, Proposition 9.16, with its ambient convention `0 ≤ ρ` made explicit.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformSign_twoSixHypercontractive : FABL.IsHypercontractiveRandomVariable 2 6 FABL.twoSixNoiseRoot FABL.signValue
theorem FABL.uniformSign_twoSixHypercontractive : FABL.IsHypercontractiveRandomVariable 2 6 FABL.twoSixNoiseRoot FABL.signValue
O'Donnell, Proposition 9.16, hypercontractive-random-variable form.
-
FABL.oneBit_twoQ[complete] -
FABL.oneBit_twoQ_sharp[complete] -
FABL.oneBit_twoTop[complete] -
FABL.uniformSign_twoQHypercontractive[complete] -
FABL.uniformSign_twoTopHypercontractive[complete]
Theorem 9.17. Let x be a uniform sign and let
q\in(2,\infty]. If
0\le\rho\le\frac1{\sqrt{q-1}},
then for all a,b\in\mathbb R,
\|a+\rho bx\|_q\le\|a+bx\|_2.
Lean code for Theorem9.3.5●5 theorems
Associated Lean declarations
-
FABL.oneBit_twoQ[complete]
-
FABL.oneBit_twoQ_sharp[complete]
-
FABL.oneBit_twoTop[complete]
-
FABL.uniformSign_twoQHypercontractive[complete]
-
FABL.uniformSign_twoTopHypercontractive[complete]
-
FABL.oneBit_twoQ[complete] -
FABL.oneBit_twoQ_sharp[complete] -
FABL.oneBit_twoTop[complete] -
FABL.uniformSign_twoQHypercontractive[complete] -
FABL.uniformSign_twoTopHypercontractive[complete]
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_twoQ (q a b ρ : ℝ) (hq : 2 < q) (hρ0 : 0 ≤ ρ) (hρ : ρ ≤ 1 / √(q - 1)) : (FABL.uniformLpNorm q fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
theorem FABL.oneBit_twoQ (q a b ρ : ℝ) (hq : 2 < q) (hρ0 : 0 ≤ ρ) (hρ : ρ ≤ 1 / √(q - 1)) : (FABL.uniformLpNorm q fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
O'Donnell, Theorem 9.17, for every finite real exponent `q > 2`.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_twoQ_sharp (q a b : ℝ) (hq : 2 < q) : (FABL.uniformLpNorm q fun x => a + 1 / √(q - 1) * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
theorem FABL.oneBit_twoQ_sharp (q a b : ℝ) (hq : 2 < q) : (FABL.uniformLpNorm q fun x => a + 1 / √(q - 1) * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
The sharp finite-`q` endpoint of Theorem 9.17.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_twoTop (a b : ℝ) : (FABL.uniformLInfinityNorm fun x => a + 0 * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
theorem FABL.oneBit_twoTop (a b : ℝ) : (FABL.uniformLInfinityNorm fun x => a + 0 * b * FABL.signValue x) ≤ FABL.uniformLpNorm 2 fun x => a + b * FABL.signValue x
The `q = ∞` endpoint of Theorem 9.17. The sharp correlation is zero.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformSign_twoQHypercontractive (q : NNReal) (hq : 2 < ↑q) : FABL.IsHypercontractiveRandomVariable 2 (↑q) (1 / √(↑q - 1)) FABL.signValue
theorem FABL.uniformSign_twoQHypercontractive (q : NNReal) (hq : 2 < ↑q) : FABL.IsHypercontractiveRandomVariable 2 (↑q) (1 / √(↑q - 1)) FABL.signValue
The finite-exponent part of Theorem 9.17 in the hypercontractive-random-variable interface.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformSign_twoTopHypercontractive : FABL.IsHypercontractiveRandomVariable 2 ⊤ 0 FABL.signValue
theorem FABL.uniformSign_twoTopHypercontractive : FABL.IsHypercontractiveRandomVariable 2 ⊤ 0 FABL.signValue
The top-exponent endpoint in Definition 9.13.
-
FABL.normalized_pTwo_affine[complete] -
FABL.affine_pTwo_hanner[complete] -
FABL.oneBit_pTwo[complete] -
FABL.oneBit_pTwo_sharp[complete] -
FABL.uniformSign_pTwoHypercontractive[complete]
Theorem 9.18. Let x be a uniform sign and let 1\le p<2. If
0\le\rho\le\sqrt{p-1},
then for all a,b\in\mathbb R,
\|a+\rho bx\|_2\le\|a+bx\|_p.
Lean code for Theorem9.3.6●5 theorems
Associated Lean declarations
-
FABL.normalized_pTwo_affine[complete]
-
FABL.affine_pTwo_hanner[complete]
-
FABL.oneBit_pTwo[complete]
-
FABL.oneBit_pTwo_sharp[complete]
-
FABL.uniformSign_pTwoHypercontractive[complete]
-
FABL.normalized_pTwo_affine[complete] -
FABL.affine_pTwo_hanner[complete] -
FABL.oneBit_pTwo[complete] -
FABL.oneBit_pTwo_sharp[complete] -
FABL.uniformSign_pTwoHypercontractive[complete]
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.normalized_pTwo_affine (p e : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) (he : |e| ≤ 1) : √(1 + (p - 1) * e ^ 2) ≤ (((1 + e) ^ p + (1 - e) ^ p) / 2) ^ p⁻¹
theorem FABL.normalized_pTwo_affine (p e : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) (he : |e| ≤ 1) : √(1 + (p - 1) * e ^ 2) ≤ (((1 + e) ^ p + (1 - e) ^ p) / 2) ^ p⁻¹
Normalized two-point form of the `(p,2)` inequality.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.affine_pTwo_hanner (p a b : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) : √(a ^ 2 + (p - 1) * b ^ 2) ≤ ((|a + b| ^ p + |a - b| ^ p) / 2) ^ p⁻¹
theorem FABL.affine_pTwo_hanner (p a b : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) : √(a ^ 2 + (p - 1) * b ^ 2) ≤ ((|a + b| ^ p + |a - b| ^ p) / 2) ^ p⁻¹
The sharp scalar two-point Hanner inequality underlying Theorem 9.18.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_pTwo (p a b ρ : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) (hρ0 : 0 ≤ ρ) (hρ : ρ ≤ √(p - 1)) : (FABL.uniformLpNorm 2 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm p fun x => a + b * FABL.signValue x
theorem FABL.oneBit_pTwo (p a b ρ : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) (hρ0 : 0 ≤ ρ) (hρ : ρ ≤ √(p - 1)) : (FABL.uniformLpNorm 2 fun x => a + ρ * b * FABL.signValue x) ≤ FABL.uniformLpNorm p fun x => a + b * FABL.signValue x
O'Donnell, Theorem 9.18, including every smaller admissible correlation.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.oneBit_pTwo_sharp (p a b : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) : (FABL.uniformLpNorm 2 fun x => a + √(p - 1) * b * FABL.signValue x) ≤ FABL.uniformLpNorm p fun x => a + b * FABL.signValue x
theorem FABL.oneBit_pTwo_sharp (p a b : ℝ) (hp1 : 1 ≤ p) (hp2 : p ≤ 2) : (FABL.uniformLpNorm 2 fun x => a + √(p - 1) * b * FABL.signValue x) ≤ FABL.uniformLpNorm p fun x => a + b * FABL.signValue x
Endpoint form of Theorem 9.18.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformSign_pTwoHypercontractive (p : NNReal) (hp1 : 1 ≤ ↑p) (hp2 : ↑p ≤ 2) : FABL.IsHypercontractiveRandomVariable (↑p) 2 (√(↑p - 1)) FABL.signValue
theorem FABL.uniformSign_pTwoHypercontractive (p : NNReal) (hp1 : 1 ≤ ↑p) (hp2 : ↑p ≤ 2) : FABL.IsHypercontractiveRandomVariable (↑p) 2 (√(↑p - 1)) FABL.signValue
Theorem 9.18 in the hypercontractive-random-variable interface.
-
FABL.uniformInner_le_uniformLpNorm_mul[complete] -
FABL.exists_uniformLpNorm_dual[complete] -
FABL.selfAdjoint_dual_bound[complete]
Proposition 9.19. Let T be self-adjoint on
L^2(\Omega,\pi). Let p' and q' be the Hölder conjugates of
p and q. If
\|Tf\|_q\le C\|f\|_p
\quad\text{for every }f,
then
\|Tg\|_{p'}\le C\|g\|_{q'}
\quad\text{for every }g.
Lean code for Theorem9.3.7●3 theorems
Associated Lean declarations
-
FABL.uniformInner_le_uniformLpNorm_mul[complete]
-
FABL.exists_uniformLpNorm_dual[complete]
-
FABL.selfAdjoint_dual_bound[complete]
-
FABL.uniformInner_le_uniformLpNorm_mul[complete] -
FABL.exists_uniformLpNorm_dual[complete] -
FABL.selfAdjoint_dual_bound[complete]
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.uniformInner_le_uniformLpNorm_mul.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] {p q : ℝ} (hpq : p.HolderConjugate q) (f g : Ω → ℝ) : FABL.uniformInner f g ≤ FABL.uniformLpNorm p f * FABL.uniformLpNorm q g
theorem FABL.uniformInner_le_uniformLpNorm_mul.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] {p q : ℝ} (hpq : p.HolderConjugate q) (f g : Ω → ℝ) : FABL.uniformInner f g ≤ FABL.uniformLpNorm p f * FABL.uniformLpNorm q g
Hölder's inequality for normalized finite-uniform norms at arbitrary finite conjugate exponents.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.exists_uniformLpNorm_dual.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q p : ℝ) (hpq : q.HolderConjugate p) (f : Ω → ℝ) : ∃ g, FABL.uniformLpNorm p g ≤ 1 ∧ FABL.uniformInner f g = FABL.uniformLpNorm q f
theorem FABL.exists_uniformLpNorm_dual.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (q p : ℝ) (hpq : q.HolderConjugate p) (f : Ω → ℝ) : ∃ g, FABL.uniformLpNorm p g ≤ 1 ∧ FABL.uniformInner f g = FABL.uniformLpNorm q f
Sharpness of normalized Hölder on a nonempty finite uniform space: the `L^q` norm has a dual witness of `L^p` norm at most one.
-
theoremdefined in FABL/Chapter09/OneBitHypercontractivity.leancomplete
theorem FABL.selfAdjoint_dual_bound.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (p q p' q' C : ℝ) (hp : p.HolderConjugate p') (hq : q.HolderConjugate q') (T : (Ω → ℝ) → Ω → ℝ) (hself : ∀ (f g : Ω → ℝ), FABL.uniformInner (T f) g = FABL.uniformInner f (T g)) (hbound : ∀ (f : Ω → ℝ), FABL.uniformLpNorm q (T f) ≤ C * FABL.uniformLpNorm p f) (g : Ω → ℝ) : FABL.uniformLpNorm p' (T g) ≤ C * FABL.uniformLpNorm q' g
theorem FABL.selfAdjoint_dual_bound.{u_1} {Ω : Type u_1} [Fintype Ω] [Nonempty Ω] (p q p' q' C : ℝ) (hp : p.HolderConjugate p') (hq : q.HolderConjugate q') (T : (Ω → ℝ) → Ω → ℝ) (hself : ∀ (f g : Ω → ℝ), FABL.uniformInner (T f) g = FABL.uniformInner f (T g)) (hbound : ∀ (f : Ω → ℝ), FABL.uniformLpNorm q (T f) ≤ C * FABL.uniformLpNorm p f) (g : Ω → ℝ) : FABL.uniformLpNorm p' (T g) ≤ C * FABL.uniformLpNorm q' g
O'Donnell, Proposition 9.19, for a nonempty finite uniform probability space.