Conditional Jensen's inequality
ProvedFamousTheorems.conditional_jensen_inequalityThe conditional Jensen inequality. Let be a convex, lower semicontinuous function on a closed convex set in a real Banach space . Let be integrable with values in almost everywhere, with integrable, and let be a sub--algebra. Then almost everywhere
This is the conditional form of Jensen's inequality. It implies that convex functions of martingales are submartingales, that conditional expectation contracts norms, and many inequalities of martingale theory.
Formalization note. Mathlib's ConvexOn.map_condExp_le. The sub--algebra is m ≤ mα, and the measure is assumed -finite on it (SigmaFinite (μ.trim hm)). μ[f | m] is the conditional expectation, and ≤ᵐ[μ] is inequality almost everywhere.
Preamble
import Mathlib
Formal statement
namespace FamousTheorems
open MeasureTheory
theorem conditional_jensen_inequality {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E] [CompleteSpace E] {α : Type*} {f : α → E} {φ : E → ℝ}
{m mα : MeasurableSpace α} {μ : Measure α} {s : Set E} (hm : m ≤ mα) [SigmaFinite (μ.trim hm)]
(hφ : ConvexOn ℝ s φ) (hφc : LowerSemicontinuousOn φ s) (hfs : ∀ᵐ a ∂μ, f a ∈ s) (hs : IsClosed s)
(hf : Integrable f μ) (hφf : Integrable (φ ∘ f) μ) :
φ ∘ μ[f | m] ≤ᵐ[μ] μ[φ ∘ f | m] := by sorry
end FamousTheoremsSource
Listed in Mathlib's curated theorem manifests (docs/1000.yaml); formalized in Mathlib as `ConvexOn.map_condExp_le`. Proof here reduces to that Mathlib result.