Stochastic kernels contract arbitrary finite signed masses
ProvedAFT.FiniteDist.signed_kernel_l1_contractionfinite-probabilityinfo-loss-lar-native-5c3becbstoc-2027
Let be any signed mass on a finite set, and let be a probability law on a finite set for every . Then
This contraction applies before clipping, without requiring the input mass to be nonnegative or normalized.
Preamble
import Definitions.Def_InfoLossLAR_Native_AFT_TotalVariation
import Mathlib
/-! Exact finite total variation and contraction by a stochastic kernel. -/
namespace AFT
end AFT
section AFT
open AFT
open scoped BigOperators Classical
variable {α β : Type*} [Fintype α] [Fintype β]
namespace AFT.FiniteDist
end AFT.FiniteDist
section FiniteDist
open AFT
open AFT.FiniteDist
Formal statement
theorem AFT.FiniteDist.signed_kernel_l1_contraction (weight : α → ℝ) (kernel : α → FiniteDist β) :
(∑ output, |∑ input, weight input * (kernel input).weight output|) ≤
∑ input, |weight input| := by sorrySource
Local manuscript repository commit 5c3becb254daaa8738368b474349766c1769c2a0; lean/AFT/TotalVariation.lean; lines 121–136; retained native proof and type checked against the native theorem.