2403.07511
THE ENTROPY OF AN EXTENDED MAP FOR ABELIAN GROUP ACTIONS
Yuan Lian
correctmedium confidence
- Category
- Not specified
- Journal tier
- Note/Short/Other
- Processed
- Sep 28, 2025, 12:56 AM
- arXiv Links
- Abstract ↗PDF ↗
Audit review
The paper proves h_μ(G, Y | Π_X) = 0 for the natural extension by approximating arbitrary partitions using sets of the form Π_{g,X}^{-1}(B) and a standard small-symmetric-difference-to-small-conditional-entropy lemma, yielding zero relative entropy along Følner sets . The candidate solution instead observes that for the constructed natural extension one has Π_{g,X}^{-1}(A) = Π_X^{-1}(α_g(A)) and hence the σ-algebra on Y is the μ-completion of Π_X^{-1}B_X; therefore every partition is (mod μ) measurable with respect to the conditioning σ-algebra, making all conditional entropies vanish immediately. Both arguments yield the same conclusion. Minor issues in the candidate (a t=g vs. g=e substitution) do not affect correctness. The paper’s proof is sound though longer than necessary; it implicitly contains the key identification Π_{g,X}^{-1}(B) ⊆ Π_{e,X}^{-1}(B) for all g, which already forces the generated σ-algebra to be Π_X^{-1}B_X (hence conditional entropy zero) .
Referee report (LaTeX)
\textbf{Recommendation:} major revisions \textbf{Journal Tier:} note/short/other \textbf{Justification:} The main result is valid but in this setup the natural extension is measurably isomorphic to the original system via Π\_X, making the theorem an immediate corollary of basic facts about conditional entropy. The present proof, while correct, is longer than necessary and obscures this simplicity. Clarifying notation and explicitly identifying the isomorphism would considerably improve the paper and its contribution.