Finite entropy and channel laws for route-sensitive context lifting
ProvedTranscendenceTower.InformationTheory.finite_route_informationIn the route-sensitive lifting example, the actual minimal routed blocking sets are exactly and . The direct transfer has the unique empty blocking set. After explicitly imposing the symmetric distribution on the two routed alternatives and a uniform prior over direct versus routed transfer, the resulting finite experiment has the displayed entropy and information values.
The proof has five parts.
- It classifies every routed minimal block, excluding both additional alternatives and irrelevant extra members.
- It identifies the direct resolution as a singleton Dirac law with zero entropy and proves that the symmetric routed law is normalized, swap-invariant, and has nats of entropy—one bit.
- It constructs and normalizes the four-cell joint distribution of the route and the Boolean survival output .
- From that joint law it derives survival probability , conditional entropy , and the stated mutual information.
- In a separate abstract loss-only model, it proves that fresh-stage memoryless channels with survival probabilities compose to survival probability ; two fair half-survival stages therefore retain the fact with probability .
Formalization note: entropy and mutual information are finite sums over explicitly normalized real-valued masses. Probability is imposed as additional modeling structure; it is not inferred from logical nondeterminism.
The result concerns this finite propositional lifting witness and a separate memoryless loss-channel abstraction. It does not establish a general probabilistic semantics for nonmonotonic logic or a Shannon coding theorem.
import Definitions.Def_TranscendenceTowerInformationTheory set_option autoImplicit false open scoped BigOperators open TranscendenceTower TranscendenceTower.ContextTheories TranscendenceTower.Lifting open TranscendenceTower.InformationTheoryPlatform
theorem TranscendenceTower.InformationTheory.finite_route_information :
((∀ A : Set L0, MinBlock srcPQ tgtNand A ↔ A = {pA} ∨ A = {qA})
∧ (∀ A : Set L0, MinBlock srcPQ tgtNand A ↔
∃ r : RoutedResolution, A = routedBlock r))
∧ ((∀ A : Set L0, MinBlock srcP tgtNand A ↔ A = ∅)
∧ IsFiniteDistribution directSingletonMass
∧ finiteEntropy directSingletonMass = 0)
∧ (IsFiniteDistribution uniformRoutedSelectionMass
∧ (∀ r, uniformRoutedSelectionMass r =
uniformRoutedSelectionMass (swapResolution r))
∧ finiteEntropy uniformRoutedSelectionMass = Real.log 2)
∧ ((∀ r output, 0 ≤ jointMass r output)
∧ (∑ r : Route, ∑ output : Bool, jointMass r output = 1)
∧ outputMarginal true = 3 / 4
∧ uniformRouteConditionalEntropy = Real.log 2 / 2
∧ uniformRouteMutualInformation = Real.binEntropy (3 / 4) - Real.log 2 / 2)
∧ ((∀ s t : ℝ, 0 ≤ s → s ≤ 1 → 0 ≤ t → t ≤ 1 →
composeLossChannel s t = lossChannel (s * t)
∧ ∀ input, IsFiniteDistribution (composeLossChannel s t input))
∧ composeLossChannel (1 / 2) (1 / 2) true true = 1 / 4) := by sorry