Entropy inequality for a finite latent observation mixture
ProvedZeroFreeRRG.Hardcore.latent_mixture_entropy_superadditivityLet have positive sum , and let be nonnegative integrable densities of mean one under a probability measure . Define and . If , , and are integrable under , then
This is the conditional relative-entropy step in the vertex entropy convexity argument.
Preamble
import Mathlib open scoped BigOperators open MeasureTheory
Formal statement
namespace ZeroFreeRRG.Hardcore
theorem latent_mixture_entropy_superadditivity {ι X : Type} [Fintype ι] [MeasurableSpace X]
(μ : Measure X) [IsProbabilityMeasure μ] (A : ι → ℝ) (w : ι → X → ℝ)
(hA : ∀ i, 0 ≤ A i) (hS : 0 < ∑ i, A i)
(hw : ∀ i x, 0 ≤ w i x) (hint : ∀ i, Integrable (w i) μ)
(hmean : ∀ i, (∫ x, w i x ∂μ) = 1)
(hTlog : Integrable (fun p : X × X =>
(∑ i, A i*w i p.1*w i p.2)*Real.log (∑ i, A i*w i p.1*w i p.2)) (μ.prod μ))
(hcrossL : Integrable (fun p : X × X =>
(∑ i, A i*w i p.1*w i p.2)*Real.log (∑ i, A i*w i p.1)) (μ.prod μ))
(hcrossR : Integrable (fun p : X × X =>
(∑ i, A i*w i p.1*w i p.2)*Real.log (∑ i, A i*w i p.2)) (μ.prod μ)) :
2*(∫ x, (∑ i, A i*w i x)*Real.log (∑ i, A i*w i x) ∂μ) ≤
(∫ p : X × X, (∑ i, A i*w i p.1*w i p.2)*Real.log (∑ i, A i*w i p.1*w i p.2) ∂(μ.prod μ)) +
(∑ i, A i)*Real.log (∑ i, A i) := by sorry
end ZeroFreeRRG.HardcoreSource
Chen, Wang, and Zhang, Zero-Freeness for Random Regular Graphs up to the Replica Symmetry Breaking Threshold, https://github.com/sigact-stoc/stoc2027/releases/download/papers-2/1461.pdf, proof of Lemma 4.27, printed p. 34.