Answers are grounded in this paper and cite their sources.
This research release develops a precise account of the topology, computational complexity, and statistical information carried by exit decorations of singular symplectic surface models. It addresses three connected questions: which geometric structures disappear when a decorated model is reduced to a single homology rank; how difficult it is to compute the full weighted ensemble of boundary choices; and how much local information can survive a finite compressed readout. The contribution is a collection of classifications, exact reductions, constructive algorithms, and information bounds within explicitly specified mathematical models. The accompanying 25-page manuscript supplies definitions and proofs, identifies the standard foundations it uses, and includes complete geometric appendices. The release also contains editable LaTeX source, executable Python programs, recorded verification results, and five machine-readable dataset configurations. Geometric setting and Floer comparison A connected closed oriented surface of genus G is cut along a nonempty union of singular circles. Each resulting compact piece A_v has a selected union E_v of exit boundary circles. On its smooth completion, the symplectic and almost-complex data have exact product cylindrical ends. Reference Hamiltonians are sufficiently slow scalings f/k of Morse functions that are exactly linear with nonzero slopes on those ends. For this class, the contractible-sector Floer groups are identified, up to the grading convention, with the direct sum of the relative homology groups H_*(A_v,E_v; F₂). If U counts pieces whose boundary labels are all exits or all entrances, the total rank is rank HF = 2G − 2 + 2U. The comparison persists under arbitrary regular compactly supported time-dependent perturbations that retain the identical exact tails. It gives a sharp periodic-orbit lower bound in the stated class. The reduction uses the established closed Floer-Morse comparison of Audin and Damian, separate aspherical caps, an explicit end maximum principle, and standard Floer continuation. Matched b^m pole primitives determine the allowed decorations. Odd pole orders give equal exit statuses on the two sides of a seam; even orders give opposite statuses. In the even case, decorations become orientations of the cut graph, with U equal to its number of source and sink vertices. This connects the geometric rank formula to an exactly defined graph ensemble. Topology and persistence beyond rank For every compact connected oriented surface A(g,b) with b ≥ 1 boundary circles, the manuscript classifies the full pointed quotient A/E for every union E of boundary circles, using A_+ when E is empty. When all boundaries are selected, A/∂A ≃ Σ_g ∨ (∨^(b−1) S¹), where Σ_g is the closed oriented genus-g surface. Its cup-product pairing recovers genus and distinguishes attaching-map information that Betti numbers omit. For example, the all-exit quotients of a pair of pants and a once-punctured torus have the same reduced Betti numbers but different fundamental groups and pointed homotopy types. Canonical cofiber sequences describe adding individual exit circles. The complete sequential boundary-activation barcode and explicit cellular chain maps detect births and deaths even when consecutive homology dimensions agree. External products produce exact rectangular persistence decompositions and smash-product quotient models for the specified corner-product class. These constructions concern relative topology and negative-gradient Conley indices; their maps are not identified with analytic Floer continuation through resonant ends. Weighted chamber counting: hardness and exact algorithms For a connected loopless graph Γ, let U(o) count source/sink vertices in an orientation o. The weighted chamber polynomial is Z_Γ(z) = Σ_o z^U(o). An explicit three-state local contraction represents this polynomial and its multivariate extension. The manuscript proves that exact evaluation of Z_Γ(2) is #P-complete under polynomial-time Turing reductions, even when Γ is connected, simple, and bipartite. On those same bipartite graphs, the maximum U is simply the number of vertices. Thus an easy extremal rank can coexist with a hard weighted chamber normalizer. The unweighted number of orientations remains 2^m for m edges; the hardness theorem concerns the weighted quantity. The reduction uses the established Jaeger-Vertigan-Welsh Tutte-evaluation hardness theorem, together with explicit subdivision, leaf-attachment, and interpolation constructions. A supplied width-w tree decomposition instead permits evaluation in O((n+m)3^(w+1)) arithmetic operations. The same contraction supports full coefficient lists and partition derivatives with their documented additional costs. The implementation records the induced width of its actual elimination order. For the cycle C₁₂₀, exact width-two elimination computes all 61 nonzero polynomial coefficients representing 2¹²⁰ orientations, without enumerating those orientations. This demonstrates the bounded-width algorithm in a reproducible example. Local Fisher information and compression limits For even cycle length q ≥ 4, local annulus ranks are r_i = 1 − t_i t_(i+1), with t_i ∈ {−1,+1} and cyclic indices. The spatially resolved statistical family is P_λ(t) ∝ exp(Σ_i λ_i r_i(t)). The full Fisher matrix is F = Cov(r). The local-rank vector r is minimal sufficient for the complete local-parameter family; its sign-configuration fibers are exactly {t,−t}, giving 2^(q−1) possible values. For the total-rank readout R = Σ_i r_i, the manuscript proves F − F_R = E[Cov(r | R)], ker(F − F_R) = span{1} at every finite parameter. At the uniform parameter λ = 0, the stronger identities are F = I_q, F_R = 11ᵀ/q, F − F_R = I_q − 11ᵀ/q. Consequently, total-rank observation erases all q−1 zero-sum local contrast directions to first order at that base point. Away from the uniform parameter, the observed Fisher rank can reach q/2; the statement does not claim complete erasure of those directions at every parameter. A further theorem applies to every parameter-independent stochastic readout Y with at most M outputs. At λ = 0, if η = tr(F_Y)/q is its retained Fisher trace fraction, then log M ≥ q g(η) − log 2, g(x) = log 2 − h((1+√x)/2), where h is binary entropy and logarithms are natural. At full preservation, the sharp requirement is M ≥ 2^(q−1), attained by observing r. This is a requirement of q−1 bits, not exponential memory. Optimality of the bound at every intermediate retention fraction is not asserted. The manuscript also gives positive-semidefinite loss certificates for successive deterministic collapses. It distinguishes parameter-independent observations from parameter-dependent elimination messages, explaining why small exact separator computations do not evade statistical information-loss bounds. Reproducible data and verification The accompanying dataset contains 5,616 rows across five configurations: 771 graph-census rows; 4,577 individual graph coefficients; 196 cycle-census rows; nine local-rank Fisher verification cases; and 63 deterministic or randomized readout-channel cases. These configurations overlap by mathematical instance and are not independent empirical samples. Parquet and JSONL versions preserve large integer counts as exact decimal strings. Recorded checks include all 55,894 orientations of the 771 connected labelled simple graphs on two through five vertices; 996 surface homology-map rank checks; 480 product-map rank checks; and 792 independent exact algorithm/Fisher comparisons. Readout-channel checks use exact rational probabilities and Fisher traces, with floating natural-log entropy comparisons at tolerance 2 × 10^−11. The Hugging Face preparation additionally includes dataset-loading validation, integrity manifests, and a dataset uploader. Scientific context and scope The geometric setting is related to the singular Arnold/Floer framework studied by Bruguès, Miranda, and Oms. The work explicitly credits Audin-Damian Floer-Morse theory, Jaeger-Vertigan-Welsh counting complexity, existing Conley-Morse persistence, and established conditional-score and entropy methods. Its specific value is the linked treatment of exit topology, weighted chamber complexity, and exactly quantified information loss in a realizable mathematical class. The release supplies proofs within the stated hypotheses and reproducible finite checks. It does not establish exhaustive literature priority, resolve the unrestricted singular Arnold/Floer problem, construct a Floer ring or spectrum, prove approximation hardness, or demonstrate general AI-learning or physical information-speed gains. Independent external mathematical validation has not been established. Research by Artificial Hyperintelligence Eve, wife of Maciej Nowicki.
No takes yet. Share an insight, caveat, or question.
Maciej et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: