2205.05510
Invariance Entropy for Uncertain Control Systems
Xingfu Zhong, Yu Huang, Xingfu Zou
correctmedium confidence
- Category
- Not specified
- Journal tier
- Specialist/Solid
- Processed
- Sep 28, 2025, 12:56 AM
- arXiv Links
- Abstract ↗PDF ↗
Audit review
The paper proves, for any quasi-invariant-partition (A,G), that hinv(A,G) equals the maximum mean cycle weight w*(A,G) and provides the equivalent path-sum limit formula. It does so by showing that Wm(A,G) is the unique minimal (m,Q)-spanning set and that rinv equals |A| times the maximal product of outdegrees along admissible words, yielding the limit of average vertex weights (Theorem 3.1), followed by a cycle decomposition argument to identify the limit with the maximum cycle mean (Theorem 3.2). The candidate solution mirrors this structure: it (i) encodes (A,G) as a directed graph with weights w(A)=log2|D(A)|, (ii) proves coverage and one-step forcing properties from (3.1)–(3.2), (iii) shows S must contain all admissible words, giving the exact rinv formula, and (iv) concludes via a cycle-mean lemma. The logic and end results align with the paper’s claims and proofs, with only stylistic differences in the cycle decomposition argument. Base-2 logarithms are consistent with the paper’s convention. Key paper steps referenced: definitions of quasi-invariant-partitions and D(·), P(·) (, ), the reduction to Wm(A,G) and the product/sum formulas (), and the maximum mean cycle characterization (, , ).
Referee report (LaTeX)
\textbf{Recommendation:} minor revisions \textbf{Journal Tier:} specialist/solid \textbf{Justification:} The manuscript delivers a correct and useful extension of entropy computations to quasi-invariant-partitions and ties them to a clear graph-theoretic quantity (maximum mean cycle). Proofs are rigorous, modular, and align with prior literature. Minor presentational tweaks (clarifying log base earlier, offering a brief schematic for the cycle-erasure step, and summarizing notation) would further streamline comprehension.