Keyboard shortcuts

Press or to navigate between chapters

Press ? to show this help

Press Esc to hide this help

Dual Repair Entropy Decomposition

Abstract

Canonical predictive-interior and forgetting-closure quotients split conditional entropy into the two repair costs.

Theorem 1.1 (Conditional entropy telescopes across both canonical repairs).

Proof. Machine-checked in Lean as D5/S3/Entropy/Observation/DualRepairEntropyDecomposition.dual_repair_entropy_decomposition (✓ std3). ∎

Source. Repository-derived.

Commentary.

Let X be finite, let F update X, let R be an equivalence relation, and let mu be a normalized law with positive mass at every state. The predictive interior and forgetting closure are the imported canonical congruence repairs.

Their inclusion proofs induce canonical quotient maps from X/I to X/R and from X/R to X/C. The displayed laws are deterministic pushforwards of mu along these quotient maps, so no entropy target is used to define a source object.

Applying the repository quotient-fiber entropy decomposition to I to C, I to R, and R to C gives three entropy balances. Pushforward composition identifies the two closure laws, and the balances telescope to the claimed equality.

Pinned Mathlib and installed-package searches found no finite real-valued conditional-entropy theorem with these canonical repair quotients.

References