Doubly stochastic holder
doubly_stochastic_holder
Plain-language statement
Doubly stochastic Hölder inequality: for nonneg a, b, doubly stochastic w, and conjugate p, q > 1: ∑{ij} a_i * b_j * w{ij} ≤ (∑ a_i^p)^{1/p} * (∑ b_j^q)^{1/q}.
Exact Lean statement
lemma doubly_stochastic_holder (a b : d → ℝ) (w : d → d → ℝ)
(ha : ∀ i, 0 ≤ a i) (hb : ∀ j, 0 ≤ b j)
(hw : ∀ i j, 0 ≤ w i j)
(hrow : ∀ i, ∑ j, w i j = 1) (hcol : ∀ j, ∑ i, w i j = 1)
(p q : ℝ) (hp : 1 < p) (hpq : 1/p + 1/q = 1) :
∑ i, ∑ j, a i * b j * w i j ≤ (∑ i, a i ^ p) ^ (1/p) * (∑ j, b j ^ q) ^ (1/q)Formal artifact
Lean source
lemma doubly_stochastic_holder (a b : d → ℝ) (w : d → d → ℝ) (ha : ∀ i, 0 ≤ a i) (hb : ∀ j, 0 ≤ b j) (hw : ∀ i j, 0 ≤ w i j) (hrow : ∀ i, ∑ j, w i j = 1) (hcol : ∀ j, ∑ i, w i j = 1) (p q : ℝ) (hp : 1 < p) (hpq : 1/p + 1/q = 1) : ∑ i, ∑ j, a i * b j * w i j ≤ (∑ i, a i ^ p) ^ (1/p) * (∑ j, b j ^ q) ^ (1/q) := by by_contra h_contra contrapose! h_contra with h_contra simp_all [ ← Finset.mul_sum, mul_assoc ] have h0q : 0 < q := lt_of_le_of_ne ( le_of_not_gt fun hq => by { rw [ inv_eq_one_div, div_eq_mul_inv ] at hpq; nlinarith [ inv_mul_cancel₀ ( by linarith : ( p : ℝ ) ≠ 0 ), inv_lt_zero.2 hq ] } ) (by grind) -- Apply Hölder's inequality with the sequences $a_i$ and $g_i = \sum_j w_{ij} b_j$. have h_holder : (∑ i, a i * (∑ j, w i j * b j)) ≤ (∑ i, a i ^ p) ^ (1 / p) * (∑ i, (∑ j, w i j * b j) ^ q) ^ (1 / q) := by have := @Real.inner_le_Lp_mul_Lq simp_all convert this Finset.univ a ( fun i => ∑ j, w i j * b j ) ( show p.HolderConjugate q from ?_ ) using 1 · simp only [ha, abs_of_nonneg, mul_eq_mul_left_iff] left congr! symm exact abs_of_nonneg (Finset.sum_nonneg fun _ _ ↦ mul_nonneg (hw _ _) (hb _)) · constructor <;> linarith -- By Fubini's theorem, we can interchange the order of summation. have h_fubini : ∑ i, (∑ j, w i j * b j) ^ q ≤ ∑ j, b j ^ q := by -- Apply Jensen's inequality to the convex function $x^q$ with weights $w_{ij}$. have h_jensen : ∀ i, (∑ j, w i j * b j) ^ q ≤ ∑ j, w i j * b j ^ q := by intro i have h_jensen : ConvexOn ℝ (Set.Ici 0) (fun x : ℝ => x ^ q) := by apply convexOn_rpow nlinarith [ inv_pos.2 ( zero_lt_one.trans hp ), inv_pos.2 h0q, mul_inv_cancel₀ h0q.ne'] convert h_jensen.map_sum_le _ _ _ <;> aesop refine' le_trans ( Finset.sum_le_sum fun i _ => h_jensen i ) _; rw [ Finset.sum_comm ] simp [ ← Finset.sum_mul, hcol] simp_all only [mul_comm]; simpa using h_holder.trans ( mul_le_mul_of_nonneg_left ( Real.rpow_le_rpow ( Finset.sum_nonneg fun _ _ => Real.rpow_nonneg ( Finset.sum_nonneg fun _ _ => mul_nonneg ( hb _ ) ( hw _ _ ) ) _ ) h_fubini (one_div_nonneg.mpr h0q.le)) ( Real.rpow_nonneg ( Finset.sum_nonneg fun _ _ => Real.rpow_nonneg ( ha _ ) _ ) _ ) ) |> le_trans <| by simp;- Project
- quantumInfo
- License
- MIT
- Commit
- 56e83a9288a3
- Source
- QuantumInfo/Finite/Entropy/Relative.lean:59-94
Reuse this declaration
Bring the exact result into your workflow
The import identifies the source module. Your project still needs the pinned package dependency shown on this page.
What this badge means
This completion status comes from the project or community source. It has not yet been represented here as an independent rebuild and axiom audit.
Continue in this project
Related declarations
Conj Transpose isometry mul isometry le one
conjTranspose_isometry_mul_isometry_le_one
Project documentation
The operator norm of the conjugate transpose is equal to the operator norm. -/ theorem Matrix.opNorm_conjTranspose_eq_opNorm {m n : Type*} [Fintype m] [Fintype n] [DecidableEq m] [DecidableEq n] (A : Matrix m n 𝕜) : Matrix.opNorm Aᴴ = Matrix.opNorm A := by unfold Matrix.opNorm rw [← ContinuousLinearMap.adjoint.norm_map (toEuclideanLin A).toContinuousLine...
Source project: quantumInfo
Person-level attribution pending.
Convex roof of pure
convex_roof_of_pure
Plain-language statement
The convex roof extension of g : KetUpToPhase d → ℝ≥0 applied to a pure state ψ is g (KetUpToPhase.mk ψ).
Source project: quantumInfo
Person-level attribution pending.
Id achieves Rate log dim
CPTPMap.id_achievesRate_log_dim
Plain-language statement
The identity channel on D dimensional space achieves a rate of log2(D).
Source project: quantumInfo
Person-level attribution pending.