Source register and mathematical mappings

This is the canonical bibliography for existing library sources and the five next README directions. Stable keys identify consulted manifestations, not interchangeable editions of a work. Dates distinguish initial public submission, version revision, online publication and issue year. Evidence was inspected on 2026-10-14. Missing evidence is labeled unresolved; neither a citation nor an accessible PDF establishes permission for every use. Errata and edition comparisons are limited to the corrections and differences explicitly recorded below; absence of another correction is not a certified exhaustive errata search. No current institutional affiliations or Japanese-name spellings are inferred from publication-era romanized attribution.

The mathematical mappings below preserve the existing digests. Source match means a selected source statement/bridge checked directly; specialization retains an explicit restricted domain; extension is a library argument beyond the attributed statement; background carries no implemented theorem claim. The bounded #215 reviews are recorded in its PRs: PH includes direct selected source comparisons; Mapper includes direct DMW/CO comparison; Reeb includes direct selected DMP comparison, with graph/PL/adaptation portions reviewed from digests. They are not complete reviews of every paper or future milestone.

Rights profiles and intended acts

The register contains metadata, links and original mathematical summaries. It does not redistribute articles, figures, supplements or third-party code, except for the previously identified software adaptation under its own license. Reading, limited quotation, AI-assisted research input/processing, local copies and repository reproduction are separate acts. Training and bulk collection are outside the requested use; no permission for them is inferred. No jurisdictional copyright exception is relied on for clearance.

Profile Explicit evidence and conditions Unresolved use/next step
AR Exact arXiv version's license link/banner identifies the nonexclusive distribution grant to arXiv. Public reading is available; arXiv license guidance distinguishes third-party reuse and version-dependent grants. This grant alone does not authorize third-party text copying, quotation, article uploads/AI processing or redistribution. For those acts establish holder permission, a separately licensed aligned manifestation, or a concrete applicable legal basis. This is not an assertion that all inference is prohibited.
CC4 Explicit article-specific CC-BY 4.0: copy/share/adapt the licensed material, with attribution, license link, preserved notices and indication of changes; no additional restrictions. Ordinary reading, licensed quotation and local copies are within those grants. Copyright permission to copy/adapt is evidence for research processing of licensed text; it is not blanket clearance for every AI service, platform contract, personal data or separately credited material. Verify those concrete conditions before an article upload. Third-party figures/code need their own grants.
U No adequate version-specific third-party reuse grant established. Availability and any notices are recorded per manifestation. Reading access is not blanket permission for quotation, local copying, article input to AI or repository reproduction. Clarify the exact holder/platform terms or use an aligned licensed alternative before claiming those permissions.
AP2 Pinned software Apache-2.0: use/reproduce/modify/distribute under its conditions, including license, notices, attribution and change notices where applicable. Applies to that code, not cited papers or unrelated datasets. The register establishes copyright-license evidence, not every possible service/security/legal condition for AI input.

Springer Nature Link terms, §§1.2–1.3, permit research-oriented access/viewing/download/printing and impose restrictions on redistribution, adaptation, systematic collection and commercial-site uploading. They do not expressly settle AI-assisted inference/research processing. For CC4 articles the article's specific grant and third-party caveats must be considered separately; no claim that generic site terms revoke CC rights is made. Subscription preview metadata is not evidence that full-text access or processing was permitted. No access controls were bypassed or new legal terms accepted in this task. Source-specific unresolved acts below concern article text/figures, not citation of bibliographic facts or original mathematical summaries.

Sources already used

Puuska17-v1 — erosion stability of rank invariants

  • Identity/version: Ville Puuska. Erosion distance for generalized persistence modules. arXiv:1710.01577v1, posted 2017-10-04; §2.2, Lemma 16, Corollary 19 and Theorem 22.
  • Use/classification: RankErosionInterleaving proves rank domination by factoring actual widened structure maps through the other module, then proves the extended-distance bound. The library compares all weak coordinate pairs, including the diagonal, for pointwise finite-dimensional field-valued modules. The source uses strict pairs; equivalence of the conventions and finite-presentation erosion attainment are not claimed by the current API. See weak-pair rank erosion.
  • Rights/evidence: AR, exact-version page license link inspected 2026-11-11. The repository supplies original proofs and mathematical summaries; no source prose or figures are reproduced.

Landi14-v1 — matching distance bounded by interleaving distance

  • Identity/version: Claudia Landi. The Rank Invariant Stability via Interleavings. arXiv:1412.3374v1, posted 2014-12-10; Lemma 1 and Theorem 1.
  • Use/classification: AffineSliceInterleaving constructs the positive-line interleaving of radius epsilon/minSlope, as in Lemma 1, and proves the extended-distance bound over arbitrary rings. AffineSliceMatching specializes Theorem 1 to original finitely presented modules over arbitrary fields, deriving the ordinary slice barcodes internally and using the same normalization max slope = 1, sum offset = 0. It proves extended pseudometric laws without separation, completeness or finite attainment claims. See affine slice matching distance.
  • Rights/evidence: U for third-party reuse; the exact primary version was read for mathematical comparison. The repository supplies original proofs and summaries, without source prose or figures.

BOO24-v3 — signed rank decompositions and restriction to lines

  • Identity/version: Magnus Bakke Botnan; Steffen Oppermann; Steve Oudot. Signed Barcodes for Multi-Parameter Persistence via Rank Decompositions and Rank-Exact Resolutions. arXiv:2107.06800v3, revised 2024-08-22; §4.3, Lemma 4.15; §5.3, Corollaries 5.8–5.9.
  • Use/classification: The source describes restriction and cancellation of signed rank decompositions; Corollary 5.8 is stated for rectangles. PositiveAffineSlice, FinitePresentationReindex, AffineSlicePresentation and AffineSliceHooks give direct library derivations of the entrance-time adjunction, original finite presentations of positive affine restrictions, and the actual finite/principal hook comparisons. Finite hook birth and death both use the maximum normalized coordinate difference. AffineSliceSignedBarcode proves the corresponding ordinary-barcode cancellation by telescoping actual rank-exact rows and comparing positive endpoint rank counts. See positive affine slices and affine slice cancellation.
  • Finite-grid Kan mapping: In §4.3 the source takes an upper semilattice P and a finite join-closed subposet Q. It describes each left Kan fiber by the value at the greatest Q-point below it, or zero if none exists; transition maps are the corresponding module map when both floors exist, and zero otherwise. It then states that the extension preserves rank-exact sequences. CoordinateGrid.lan_isRankExact gives the finite-product-grid case over a field: from an IsRankExact sequence (whose middle fibers are pointwise finite-dimensional), it derives pointwise exactness, rank additivity and pointwise finite-dimensional middle fibers for the actual Lan maps. See rank exactness and hook lifting. Arbitrary upper-semilattice cases are not formalized.
  • Rights/evidence: AR, exact-version page license link inspected 2026-10-29. The §4.3 setup and Lemma 4.15 statement were inspected in exact-version HTML on 2026-11-11; the proof was not audited. The repository provides original proofs and summaries; no source prose or figures are reproduced. The comparison is limited to the identified setup and theorem statements.

BOOS24-v3 — generated-upset projective resolutions

  • Identity/version: Magnus Bakke Botnan; Steffen Oppermann; Steve Oudot; Luis Scoccola. On the bottleneck stability of rank decompositions of multi-parameter persistence modules. Consulted arXiv:2208.00300v3, posted 2024-08-30; Proposition 5.1 and Lemmas 5.6–5.7, with §2.4, Definitions 5.4 and Notation 5.9. The page links the arXiv perpetual non-exclusive distribution license; no separate journal-version comparison is needed for these results.
  • Use: Proposition 5.1 gives the join-indexed projective resolution of the module on a finitely generated upset. The library constructs its actual signed deletion maps, proves pointwise exactness through the active-occurrence simplex, and packages them as the canonical Mathlib GeneratedUpsetResolution.projectiveResolution. The theorem generatedSupport_hasProjectiveDimensionLE gives the actual ordinary m−1 bound for arbitrary lattices, with truncated subtraction for the zero-family case; generatedSupport_projectiveDimension_le states its numerical form. Repeated labels retain distinct occurrence indices. The source's strict-join hypothesis concerns shortestness and is unnecessary for this upper bound. See join-indexed upset resolutions. Lemma 5.6 supplies the projective-resolution mechanism specialized in the derived incidence-algebra vertex tensor: the library proves the actual ideal row and projectivity, applies the actual character coinvariants, and identifies all positive left-derived values on the diagonal quotient as zero. The classification step in Lemma 5.7 is specialized in simple right modules over the diagonal quotient: actual right quotient simples are identified with their unique surviving strict-vertex characters. The quotient base-change prerequisite is realized by the actual right-module functor and adjunction in right-module base change along the diagonal quotient, including projective preservation and the canonical degree-zero vertex comparison. The finite-length reduction step is recorded in finite-length quotient dimensions. The right-module and vertex convention of §2.4, Lemma 2.1 and Remark 2.2 is formalized for the actual interval-pair incidence ring by the explicit all-vertex fiber construction in actual vertex representations of right incidence modules, including full faithfulness, the short-exactness criterion and the finite-dimensionality criterion. The vertex-representation construction itself does not establish essential surjectivity or projective-dimension transport. A separate finite-free comparison in TDA.Persistence.IncidenceVertexProjectiveDimension proves the actual finite-dimensional right-incidence and right-quotient module bounds by combining that representation with TDA.Persistence.OrdinaryFiniteLatticeDimension; it also gives the actual 2n grid bounds. This route does not claim the whole source equivalence. A further result proves the ordinary projective-dimension bound D(P) for the actual two-injective Hom cokernel of every pointwise-finite module on a finite bounded lattice, and on finite coordinate grids it proves the actual Hom-cokernel bound 2n, the actual Hom-module bound 2n−2, and the corresponding minimal-resolution termAt vanishing; see finite-grid Hom-module and Hom-cokernel dimensions. The real-coordinate degree result reuses Theorem 2.7(2)'s presentation-grid Kan reconstruction and transfers the upper-bound part of Proposition 5.17 and Theorem 5.18: every certified minimal resolution of an originally finitely presented module in n ≥ 1 real coordinates vanishes above degree 2n−2. It makes no lower-witness or equality claim and does not itself include the signed matching argument; the finite signed result is in the signed hook bottleneck overview. These results are bounded conclusions, not the global-dimension conclusion of Lemma 5.7 or a global-dimension bound for arbitrary incidence modules. See also ordinary projective dimensions of actual incidence modules.
  • Rights/evidence: AR, Proposition 5.1 inspected 2026-10-24; Lemma 5.6, Definition 5.4 and Notation 5.9 inspected 2026-10-26; §2.4 and the simple-module step of Lemma 5.7 inspected 2026-10-27; Theorem 2.7(2), Proposition 5.17 and the upper-bound part of Theorem 5.18 inspected 2026-10-27. The arXiv page links its nonexclusive distribution grant. The library documentation gives bibliographic attribution and an original mathematical summary; no article text, figures or code are reproduced.

The following manifestations each have at least one Lean consumer named in their own use line.

TauCeti — finite-length projective-dimension induction

  • Source/revision: The TauCeti Dimension.lean file at immutable commit 7b0d106bf79dbea38ba8819985bd029989181287.
  • Declaration reused: ModuleCat.hasProjectiveDimensionLT_of_isFiniteLength uses IsFiniteLength induction, the subtype/quotient short-exact row, and ShortExact.hasProjectiveDimensionLT_X₂.
  • Boundary and license: The #701 helper adapts that proof structure to the actual diagonal quotient and its #699 simple-module classification; TauCeti is not imported or added as a dependency, and no generic finite-length API is duplicated. The source header credits The Tau Ceti contributors (© 2026) and states Apache-2.0; see the revision's LICENSE. The adaptation and its boundary are also documented in finite-length quotient dimensions.

CCDS14-v1 — finite intervals and observable modules

  • Identity/version: Frédéric Chazal; William Crawley-Boevey; Vin de Silva. The observable structure of persistence modules. arXiv:1405.5644v1, initial/only submission 2014-05-22. Related publication: Homology, Homotopy and Applications 18(2), 247–265 (2016), DOI 10.4310/HHA.2016.v18.n2.a14, publication date 2016-11-29 per the Leeds institutional record. Publisher PDF retrieval failed; its license and full version comparison are unverified.
  • Printed affiliations: INRIA Saclay, GEOMETRICA; Pure Mathematics, University of Leeds; Mathematics, Pomona College, respectively (v1 closing author block).
  • Use: §1.1 finite inclusive intervals translated to half-open finite index bars; finite-interval-rank.md, finite-sublevel-diagram.md. Decomposition-free nonnegativity, finite telescoping reconstruction and terminal-tail semantics are library extensions, not the paper's observable-category theorem. Existing unversioned duplicate is resolved to the only posted v1.
  • Rights: AR, exact-version abstract license link; a separate holder notice is not established. Text quotation/copying/AI input/redistribution needs holder clarification or an aligned licensed manifestation; none verified here.

CSEH07 — persistence-diagram stability

  • Identity/version: David Cohen-Steiner; Herbert Edelsbrunner; John Harer. Stability of Persistence Diagrams. publisher article, DOI 10.1007/s00454-006-1276-5, Discrete & Computational Geometry 37, 103–120 (2007); online 2006-12-12, issue 2007-01. Earliest public manuscript date is unresolved. Consulted publisher PDF, §2 pp. 106–108; no separate numbered preprint is asserted.
  • Affiliations: Cohen-Steiner: INRIA, Sophia-Antipolis; Edelsbrunner: Computer Science, Duke University, and Geomagic, Research Triangle Park; Harer: Mathematics, Duke University (publisher PDF p. 1). No current affiliations inferred.
  • Use: source-matched persistent-rank multiplicities/triangle counts at distinct noncritical thresholds; finite-face filters, critical/self thresholds, essential tails and pairing-changing finite-filter stability are checked library specializations/extensions; see finite-sublevel-diagram.md and finite-sublevel-diagram.md §Finite-filter stability. No full general tame-function stability formalization is claimed.
  • Rights: U, publisher records ©2006 Springer and subscription preview; no CC grant established. Generic Springer research-download terms are described above; AI input/quotation/redistribution needs exact clarification or an aligned licensed alternative. No such alternative verified.

ZC04 — column operations

  • Identity/version: Afra Zomorodian; Gunnar Carlsson. Computing Persistent Homology. Consulted author-hosted SCG'04 ACM conference PDF, §4. Related journal article, DOI 10.1007/s00454-004-1146-y, Discrete & Computational Geometry 33, 249–274 (2005), online 2004-11-19. Conference DOI 10.1145/997817.997870, publication 2004-06-08 per ACM-deposited Crossref metadata; earliest public/upload date remains unverified; the conference PDF is not labeled the 2005 VoR.
  • Affiliations: Zomorodian: Computer Science, Stanford University; Carlsson: Mathematics, Stanford University (conference PDF p. 1). SCG’04 took place 2004-06-08–11, Brooklyn; event dates do not establish the upload date. No current affiliation inferred.
  • Use: source-matched field-based column operations; persistence-reduction.md. The rectangular reducer interface and reducer-to-sublevel-diagram bridges are library constructions with their own hypotheses.
  • Rights: conference PDF p. 1 ACM notice permits personal/classroom copies without profit/commercial advantage, retaining notice and full citation; republication/server redistribution requires prior permission/fee. This is narrower than a CC grant. U for AI text input and repository reproduction: clarify those specific acts with the rights holder or establish an aligned licensed alternative. Author hosting alone does not resolve them; the journal preview supplies no broad license.

Bauer21-v2 — Ripser reduction conventions

  • Identity/version: Ulrich Bauer. Ripser: efficient computation of Vietoris–Rips persistence barcodes. Consulted arXiv:1908.02518v2, 2021-02-26; initial submission 2019-08-07. Related VoR, DOI 10.1007/s41468-021-00071-5, Journal of Applied and Computational Topology 5(3), 391–423 (2021), online 2021-06-17, print 2021-09, also recorded in publisher-deposited Crossref metadata. Section/proposition references in persistence-reduction.md and ColumnReduction.lean use v2; the DOI-only older entry does not establish a switch of consulted edition.
  • Printed affiliation: Technische Universität München (v2 author footnote).
  • Use: §3.1 maximum-nonzero-row pivot/column-reduction convention, source match. Certified finite F₂ computation and homology/diagram semantics are the bounded library implementation, not a Lean implementation of all Ripser optimizations.
  • Rights: v2 AR. VoR CC4, article “Rights and permissions” with third-party-material exception and author-retained copyright; it is a prospective licensed alternative, requiring the relevant statement/numbering alignment before substitution. Ripser program's MIT license is separate and does not license the preprint.

LOT21-v4 — differential calculus

  • Identity/version: Jacob Leygonie; Steve Oudot; Ulrike Tillmann. A Framework for Differential Calculus on Persistence Barcodes. Consulted accepted manuscript arXiv:1910.00960v4, 2021-05-04; initial submission 2019-10-02. Related VoR, DOI 10.1007/s10208-021-09522-y, online 2021-07-13; Foundations of Computational Mathematics 22, 1069–1131 (2022).
  • Printed affiliations: Leygonie/Tillmann: Mathematical Institute, University of Oxford; Oudot: INRIA Saclay Île-de-France (accepted manuscript; same institutional blocks in VoR).
  • Use: selected source comparisons for ordered-barcode quotient/lift semantics, regularity and derivatives/Hessians; diagram-calculus.md, finite-diagram-ordered-q.md. Lemma 4.10/definition (9) specialize to finite order chambers; Lemma 4.6/Proposition 4.5 motivate reindexing. Explicit lift hypotheses, finite cut/rank transport and attained-grade diagram transport are library specializations/extensions. No arbitrary global differentiability or ambient-threshold naturality is claimed.
  • Rights: v4 AR; VoR CC4 (publisher “Rights and permissions”, ©The Author(s), third-party exception). The licensed VoR is an alternative for licensed text reuse, but replacing preprint theorem references requires explicit alignment; this register does not certify complete edition equality or publisher/service AI terms. No errata/full revision comparison is certified.

DMP15-v1 — Reeb/cosheaf constructions

  • Identity/version: Vin de Silva; Elizabeth Munch; Amit Patel. Categorified Reeb Graphs. Consulted arXiv:1501.04147v1, submitted 2015-01-17, internally dated 2015-01-20. Related VoR, DOI 10.1007/s00454-016-9763-9, online 2016-04-13. Full version/hypothesis comparison is not certified.
  • Affiliation evidence: no formal author-affiliation block established in v1; acknowledgments identify de Silva's home Pomona College and Munch/Patel as IMA postdoctoral fellows during the project. These are acknowledgments, not inferred printed author affiliations.
  • Use: selected source matches and corrected/specialized predicates in §2–4, detailed in the mapping below and linked Reeb digests; graph/PL adaptations retain their narrower review scope. In particular literal Definition 3.9's empty-source issue requires the corrected nonempty-source predicate for Proposition 4.19. No unproved semigroup/coherent geometric interleaving result is silently attributed. Translation of interval precosheaves and of graph objects (reeb-rgraph-translation.md) consists of library statements with §4.1–4.4 as context. #227 direction 4 is future work.
  • Rights: v1 AR; separate holder notice not established. Publisher preview identifies no CC grant; not a verified licensed alternative. Text quotation/copying/AI input/reproduction needs exact holder/platform clarification or an aligned licensed version.

DMW16-v2 — cover towers

  • Identity/version: Tamal K. Dey; Facundo Mémoli; Yusu Wang. Multiscale Mapper: Topological Summarization via Codomain Covers (consulted v2 full text). arXiv:1504.03763v2, 2016-01-12; initial submission 2015-04-15. Abstract metadata instead prints “Mutiscale Mapper: A Framework for Topological Summarization of Data and Maps”; it is not the v2 manuscript title. Related SODA proceedings DOI 10.1137/1.9781611974331.ch71: SIAM-deposited Crossref metadata records online 2015-12-21 and print 2016-01. The SIAM publisher record independently confirms online 2015-12-21, pp. 997–1013 and ©2016 SIAM, with subscription access and no explicit CC grant. Full edition alignment and specific reuse/AI terms remain unverified.
  • Printed affiliations: Dey/Wang: Computer Science and Engineering, The Ohio State University; Mémoli: Mathematics and Computer Science and Engineering, The Ohio State University (v2 p. 1).
  • Use: §2–3 indexed cover maps, pullbacks and refinement towers; source-based Mapper review. Arbitrary sets and support-aware range coverage are explicit library extensions; no metric approximation is claimed from the bare interface. §4.1 Definition 5, Proposition 4.1 and Proposition 4.2 with Corollary 4.3: interleaved cover towers give interleaved Mapper homology persistence modules, stated as a real interleaving of modules rather than of nerve towers up to contiguity. Persistence diagrams and bottleneck bounds are obtained only for step towers on finite grids with an empty initial family (zero left tail), through the BL15-v4-specialised finite-constructible isometry; this diagram statement is a library specialisation and is not attributed to a numbered DMW16-v2 result. The centred-tower perturbation bound is a library specialisation, not Proposition 4.5 or Corollary 4.6; for capped centred step towers it also has a bottleneck form, still a library specialisation and not attributed to Proposition 4.5, Corollary 4.6 or Theorems 4.7–4.8.
  • Rights: AR, v2 abstract license link/banner; separate holder notice not established. Text copying/quotation/AI input/redistribution needs holder clarification or an aligned licensed version; none verified here.

CO16-v3 — ordinary Mapper and MultiNerve

  • Identity/version: Mathieu Carrière; Steve Oudot. Structure and Stability of the 1-Dimensional Mapper. Consulted arXiv:1511.05823v3, 2016-03-29; initial submission 2015-11-18. Later arXiv v4, 2017-11-09, is labeled “Minor corrections”; the consulted §2.5/§3 statements have not been compared against v4. Related VoR title uses “One-Dimensional”; DOI 10.1007/s10208-017-9370-z, Foundations of Computational Mathematics 18(6), 1333–1396 (2018), online 2017-10-30, print 2018-12 (Crossref metadata). The arXiv journal-reference field names a different journal incorrectly; publisher metadata controls that field here.
  • Affiliations: no author-affiliation block located in v3; the later publisher record lists both at Inria Saclay, which is separate VoR evidence.
  • Use: Cover granularity and Lemma 2.10 (“no three members of a minimal cover meet”) from §2.4 are cited as background for mapper.md; the library's uniform-family multiplicity bound is proved directly and is not attributed to that lemma. §2.5 and §3 ordinary Mapper as the nerve of the path-component pullback cover, distinct MultiNerve retaining intersection-component data. Definition 2.12 (Mapper), Definition 3.1 (simplicial poset), Definition 3.2 (multinerve), Definition 3.3 (MultiNerve Mapper) and Lemma 3.4 (projection onto the nerve) are formalized for indexed families, in v3 numbering. The paper's “connected” follows its stated path-connected convention. Source-based selected Mapper comparison; no new stability/approximation theorem claimed by the present register.
  • Rights: v3 AR; holder notice not established separately. The related VoR’s Crossref entry links Springer TDM terms; its publisher preview has no CC grant; text quotation/copying/AI input/redistribution requires specific clarification or an aligned licensed alternative. No full v3/VoR revision or errata comparison is certified.

DL20-v4 — persistence measures and partial transport

  • Identity: Vincent Divol; Théo Lacombe. Understanding the Topology and the Geometry of the Space of Persistence Diagrams via Optimal Partial Transport. arXiv:1901.03048v4, related publication DOI 10.1007/s41468-020-00061-z, Journal of Applied and Computational Topology 5(1), 1–53 (2021), online 2020-10-24, print 2021-03 (publisher record, Crossref metadata). The arXiv journal reference says 2020; online publication was in 2020, while the journal issue year is 2021.
  • Manifestation/dates: corrected v4, 2024-05-28; initial arXiv submission 2019-01-10. Full text opening correction replaces the vague topology on transport-plan space by the stronger VM topology, affecting Propositions 3.1 and 3.8. Do not silently substitute the 2020 VoR for v4.
  • Printed affiliations: both Datashape, Inria Saclay (v4 opening author block).
  • Use/classification: The finite first-order matching cost uses formula (1)'s p=1, q=∞ partial-matching convention in a finite occurrence model. Attainment, label invariance, self cost, symmetry, triangle inequality, positive-lifetime separation by value-preserving occurrence bijections, and a PI corollary are proved independently. Appendix B's measure-gluing metric argument is distinct from the finite charge-allocation proof. The finite boundary-reservoir optimization is proved equal to the matching minimum through cost-preserving augmentation and mathlib Birkhoff decomposition. The real-exponent finite extension uses powered coefficients for real p≥1 at fixed q=∞ and roots after minimization; its rooted distance laws are proved by finite charge allocation and pinned mathlib Minkowski, corresponding to Proposition 3.1 / Appendix B’s metric-law mechanism. The native atomic/indexed bridge identifies these finite matching values with the infimum over all Borel boundary plans for their canonical atomic measures, including repeated unit occurrences; see the atomic comparison. The source's finite-case proof of Proposition 3.2 invokes LCO18-v2 Proposition 1. §5.1 is background for representations, not a formalized measure-level stability theorem. No Radon-measure completeness, barycenter or infinite-diagram result is claimed. Cardinality-controlled finite-part bounds and their sublevel consequences are checked library reasoning from formulas (1)/(2), full bottleneck attainment and the existing filtration stability theorem; explicit finite radii and occurrence-count budgets are retained, with essential parts excluded from the transport quantity. See finite-part perturbation bounds. The real-exponent limit to full bottleneck on fixed essential-free finite diagrams is a checked specialization of §1.1 formulas (1)/(2), proved by a finite charge-vector squeeze with the empty case separate; see the exponent limit.
  • Finite real weights: Finite real-weight boundary transport adapts Definition 2.1 and §3.2 Lemma 3.2 to arbitrary finite nonnegative real marginals, with both boundary residuals. Compact feasible matrices give attained powered/rooted optima; positive finite weighted diagrams and exact unit-mass recovery are included. The native atomic/indexed bridge proves equality with the all-Borel boundary-plan infimum for these finite weighted inputs. A weighted exponent limit remains unproved. Identification with an independently maintained ordinary Wasserstein API and comparison with the corrected article remain unresolved in #505.
  • Rights/evidence: profile AR, exact v4 abstract “view license” and full-text license banner: arXiv distribution license. Authors are identified; a separate copyright-holder notice was not established. Further text copying/AI processing/reproduction needs the clarification described in AR. The related VoR’s Crossref entries link Springer tdm/vor terms, and its publisher page identifies no CC grant; no licensed corrected alternative is verified.

LCO18-v2 — finite boundary augmentation and assignment

  • Identity: Théo Lacombe; Marco Cuturi; Steve Oudot. Large Scale computation of Means and Clusters for Persistence Diagrams using Optimal Transport. arXiv:1805.08331v2, consulted full text, 17 pages, with Supplement §7.1. The PDF identifies the 32nd Conference on Neural Information Processing Systems (NIPS 2018), Montréal.
  • Manifestation/dates: initial arXiv submission 2018-05-22; consulted v2 revision 2018-11-13. No equivalence with an independently retrieved proceedings manifestation is asserted.
  • Printed affiliations: Lacombe and Oudot: Datashape, Inria Saclay; Cuturi: Google Brain, and CREST, ENSAE (v2 opening author block).
  • Use/classification: Proposition 1 and Supplement §7.1 give the finite boundary-augmentation and assignment proof pattern for finite boundary transport. The displayed source convention ties the ground norm to p. The library's p=1, q=∞ real reservoir matrices and explicit Birkhoff reduction are a checked library adaptation of that cost-matrix argument, not the displayed theorem verbatim. No entropic regularization, solver, arbitrary-mass measure identification or barycenter theorem is claimed.
  • Rights/evidence: profile AR, exact v2 abstract's nonexclusive arXiv distribution license. No separate third-party reuse grant or holder notice was established. The AR clarification for further text copying, AI processing or redistribution remains; no separately licensed aligned manifestation is verified. This entry supplies metadata, links and an original mathematical summary.

BL15-v4 — induced matchings and isometry

  • Identity: Ulrich Bauer; Michael Lesnick. Induced Matchings and the Algebraic Stability of Persistence Barcodes. arXiv:1311.3681v4; related DOI 10.20382/jocg.v6i2a9, Journal of Computational Geometry 6(2), 162–191 (2015, arXiv journal reference; publisher retrieval failed).
  • Manifestation/dates: v4 submitted 2016-04-30, manuscript internally dated 2015-01-26; initial submission 2013-11-14. Full text. Its expanded-version notice mentions Proposition 5.10 and Remark 6.7; no full v1/VoR comparison or errata search is certified.
  • Printed affiliations: Bauer: Technische Universität München; Lesnick: Institute for Mathematics and its Applications, Minneapolis (v4 author footnotes).
  • Use/classification: Finite-constructible isometry is a checked specialization of §3 Theorem 3.5, §6 induced-matching/algebraic stability and §8.1 converse stability to finite, zero-left, right-continuous half-open modules with essential tails and finite-dimensional critical fibers over a field. It does not cover all decorated intervals or general infinite/q-tame modules. Existing finite-filter stability is a library proof and is not presented as this real-index theorem. RankSignedBarcodeOneDimensional uses the same [b,d) finite interval and [b,∞) essential-tail convention to realize the original degree-zero hook multiset as the ordinary barcode.
  • Rights/evidence: profile AR, v4 abstract license link/full-text banner. Separate holder notice not established. Exact next step: verify publisher license and manifestation alignment before treating it as an alternative for text copying/AI input; a DOI alone supplies neither.

AE17 — persistence images

  • Identity: Henry Adams; Tegan Emerson; Michael Kirby; Rachel Neville; Chris Peterson; Patrick Shipman; Sofya Chepushtanova; Eric Hanson; Francis Motta; Lori Ziegelmeier. Persistence Images: A Stable Vector Representation of Persistent Homology. JMLR 18(8), 1–35 (2017), VoR PDF.
  • Manifestation/dates: consulted JMLR article, published 2017-02 (p. 1); first arXiv submission 1507.06217 2015-07-22. The arXiv author order differs: do not transplant it into the JMLR record. No journal DOI is established here.
  • Printed affiliations: Adams, Emerson, Kirby, Neville, Peterson, Shipman: Mathematics, Colorado State University; Chepushtanova: Mathematics and Computer Science, Wilkes University; Hanson: Mathematics, Texas Christian University; Motta: Mathematics, Duke University; Ziegelmeier: Mathematics, Statistics, and Computer Science, Macalester College (article p. 1).
  • Use/classification: GaussianKernel and PersistenceImage construct the §4 Definitions 1–2 finite weighted Gaussian surface/pixel representation and prove an every-partial-matching pixel ℓ¹ estimate, with a corollary at the attained finite matching minimum. The maximum-norm Lipschitz weight hypothesis and explicit nonsharp constant are a proved library variant of the §5 route, not the paper’s exact gradient/Euclidean constant or a measure-transport identification. See persistence images. AE17 uses no arctan weight: the arctan persistence image with constant 2C + 3·√(π/2)/σ, and the composition of finite-part images with the same-label Čech bottleneck bound, are library compositions, not attributed to AE17; see arctan persistence images and Čech consumers. Landscape and scale-space-kernel precedents below concern different representations.
  • Correction: three-page erratum corrects §6.4.1 linked-twist-map equations to sequential composition; it says experiments used the correct map. It does not change the §4–5 PI stability statements. No erratum date/license is printed in the retrieved text. The linked PDF’s CreationDate and ModDate are both D:20210223183847Z (2021-02-23 18:38:47 UTC), inspected on 2026-10-14; these file-metadata timestamps do not establish a publication date. No separate reproduction permission is inferred. Its printed order is Adams, Emerson, Kirby, Peterson, Shipman, Chepushtanova, Hanson, Motta, Neville, Ziegelmeier. Changed affiliation blocks: Emerson, Data Science and Analytics Group, Pacific Northwest National Laboratory; Motta, Mathematical Sciences, Florida Atlantic University; Neville, Mathematics and Statistics, Northern Arizona University. Other institutional blocks match the article; these are erratum-era labels, not article attribution or current-affiliation claims.
  • Rights/evidence: profile CC4; article p. 1 explicitly prints CC-BY 4.0 and ©2017 Adams et al., with attribution requirements at the JMLR landing page. That notice applies to this article, not automatically its linked code/data or the separate erratum. No figures, code or supplementary files are reproduced here.

KFH18 — weighted kernel embedding

  • Identity: Genki Kusano; Kenji Fukumizu; Yasuaki Hiraoka. Kernel Method for Persistence Diagrams via Kernel Embedding and Weight Factor. JMLR 18(189), 1–41 (2018), VoR PDF. This author list is distinct from OH17-v2.
  • Manifestation/dates: JMLR article, revised 2018-01 and published 2018-04 (p. 1); initial arXiv:1706.03472 submission 2017-06-12. Volume 18 does not make its publication year 2017. No journal DOI is established here; no full preprint/VoR difference or errata search is certified.
  • Printed affiliations: Kusano: Graduate School of Science, Tohoku University; Fukumizu: The Institute of Statistical Mathematics; Hiraoka: Advanced Institute for Materials Research, Tohoku University (p. 1).
  • Use/classification: GaussianGram and WeightedKernel construct the weighted Gaussian kernel embedding of §3.1 eq. (3) for finite occurrence families and diagram finite parts in mathlib's RKHS.OfKernel. They prove the eq. (12) inner product, the §3.3 distance formula and positive semidefiniteness of the linear and eq. (13) Gaussian diagram Gram matrices. The kernel is in birth–death coordinates with the Euclidean norm, and the weight is the persistence-image Weight in birth–persistence coordinates. GaussianFeature and WeightedKernelStability formalize Lemma 16 (eqs. (25)–(26)), with Lemma 17 supplied by the existing birth–persistence bound. They prove the Proposition 10 shape under (K) and (W2) for finite partial matchings, the finite first-order cost and a count-times-bottleneck bound, using the constant L_k B₂ + B_k L₂ = √2·B/|σ| + 2A. They also prove matching-cost stability of the eq. (13) Gaussian kernel through a second-level Gaussian RKHS, which the source does not state. ArctanWeight formalizes the §3.2 w_arc weight for p = 1 as a Weight C (π/2), truncated at zero persistence below the diagonal to meet the nonnegative interface and equal to w_arc on the closed upper half-plane, and the Corollary 11 w_arc line with constant 2C + √2·π/(2|σ|) against the finite matching cost and count-times-bottleneck. RipsWeightedKernel composes these bounds with the Rips bottleneck bound, and CechWeightedKernel with the same-label Čech bottleneck bound; both compositions are library results. ArctanPersistenceImage places the w_arc weight in the AE17 persistence image with the library constant 2C + 3·√(π/2)/σ; that arctan persistence image and its Rips and Čech consumers are library compositions, not attributed to KFH18. See arctan persistence images and Čech consumers. WeightedKernelInjectivity formalizes Proposition 6 for finitely supported signed measures (the case of eq. (3)) via linear independence of Gaussian kernel functions. WeightedKernelIsometry proves the Proposition 7 multiplication isometry on the open upper half-plane, using a generic scalar PSD kernel and nowhere-zero weight; source positivity implies nonvanishing and C₀-universality is not required by the direct completed-space proof. Its Gaussian and arctan specializations allow all real bandwidths, and its finite-diagram norm comparison with the existing full-plane embedding is a library derivation. No full-plane restriction isomorphism is claimed. The generic C₀-kernel/measure-embedding equivalence underlying Proposition 6 remains unformalized. The full-plane Gaussian finite-signed-measure and C₀ case is proved separately under SFL11; its weighted positive-measure consumer allows infinite original mass with finite weighted mass. It does not assert C₀ universality for the Gaussian restricted to the open half-plane or for a second-level Gaussian kernel. Assumption (W1), Proposition 8, Theorem 9, Corollary 12 remain not formalized. See the arctan weight and Rips consumers. The §4.1.3 pixel/RKHS norm distinction is retained: no identification with PersistenceImage is claimed. The log-scale Rips-versus-Čech consumers are library compositions, not source theorems. See weighted kernel embedding.
  • Additional formalized scope: Natural-exponent total persistence and size-dependent finite-bar bounds for finite Rips and Čech diagrams are formalized. KFH18 Lemma 4's card-independent total-persistence estimate and assumption (W1) remain unformalized.
  • Native finite-measure extension: WeightedBoundaryTransportStability proves the actual full-plane RKHS bound with constant 2*A + √2*B/|σ| under native first-order boundary transport for arbitrary finite original Borel measures. The extended-valued estimate needs no moment; the real corollary and native same-mass quantization convergence use finite first diagonal moment. This is a library extension of the weighted-feature argument, not a literal KFH18 finite-diagram theorem. See Gaussian stability for finite persistence measures.
  • Rights/evidence: profile CC4; p. 1 explicitly prints ©2018 Genki Kusano, Kenji Fukumizu and Yasuaki Hiraoka, CC-BY 4.0, and JMLR attribution URL. Code, data and any third-party figures need their own evidence if later reused; none are reproduced here.

CDSO13-v3 — metric-data persistence

  • Identity: Frédéric Chazal; Vin de Silva; Steve Oudot. Persistence stability for geometric complexes. arXiv:1207.3885v3, full text.
  • Manifestation/dates: v3 submitted 2013-11-15, manuscript dated 2013-11-11; initial submission 2012-07-17. Related publisher article, DOI 10.1007/s10711-013-9937-z: Geometriae Dedicata 173, 193–214 (2014), online 2013-12-06, issue 2014-12. Full v3/VoR comparison remains unverified; the publisher preview establishes no CC grant.
  • Attribution evidence: first-page author footnotes give email addresses only, not institutional affiliations; abstract metadata assigns Chazal/Oudot to INRIA Saclay and de Silva to Pomona College. These metadata affiliations are kept distinct from printed manuscript affiliations. No current affiliations checked.
  • Use/classification: The finite metric-input construction matches the Rips diameter threshold at nonnegative scales (§4.2.1). The maximum discrepancy has the identity-correspondence distortion formula from §4.2 for a nonempty finite labelled set. The same-label sublevel-diagram estimate is an independently proved consumer of finite-face stability, not an export of Lemma 4.3 or Theorem 5.2. Singleton value zero and empty negative-time stages follow the library's support-aware convention; the source's nonpositive-scale vertex remark is not used, and no full real-indexed convention equivalence is claimed. The closed ambient Čech complex matches §4.2.3 with all of X as witnesses, and §4.2.2 for the intrinsic case; the Rips/Čech sandwich, value bounds and proper-space attainment are library extensions, since the source states no inclusion. The same-label Čech displacement bounds on values and diagrams are a library extension consistent with Theorem 5.6, since the Hausdorff distance of finite labelled images is at most the displacement; they are proved through finite-face stability and are not a Hausdorff export of Corollary 4.10 or Theorem 5.6. The witness-space comparisons between §4.2.2 and §4.2.3 complexes for nested witness sets, and the ℓ∞ equality of Čech and doubled Rips complexes, are library extensions. The log-scale Čech–Rips interleaving of homology modules with parameter log 2, and the radius-level factorizations Rips_r ⊆ Čech_r ⊆ Rips_{2r}, are library extensions built on the same Rips and §4.2.3 definitions; no source theorem is exported. The different-label ambient Čech results realize the fixed-witness relation mechanism of Lemma 4.9 and Corollary 4.10; transfer to infimum-radius filters without center attainment, arbitrary-ring interleavings, duplicate labels and full finite-diagram Hausdorff stability are library extensions. Finite ranges supply nearest points at the exact Hausdorff threshold. The ambient space stays fixed, and infinite targets appear only in finite-sample error bounds. The finite GH extension is described below. The signed face vertex-map bridge independently proves the arbitrary-map chain and homology functor used in §3 and a signed prism homotopy over arbitrary rings. The shifted natural transformation and canonical selector independence realize Proposition 3.3's global-selector/naturality/contiguity mechanism on the ordered chain model, for all real shifts. The arbitrary-ring coefficient theorem is a library generalization specializing to the source's fixed field; the explicit nonempty-face/add-only-empty convention preserves support. Native relation composition and canonical shifted homology composition realize Proposition 3.5's finite-witness and composed-selector mechanism, with explicit addition transport and the existing nonnegative-shift specialization. Numbered metric stability theorems remain separate. The Munkres proofs cited there were not directly audited for this bridge. Total-subrelation invariance realizes Proposition 3.4's shared-selector proof. The canonical interleaving from a native relation and its converse realizes §4.1 Proposition 4.2's deduction through diagonal containments; the two-relation constructor and arbitrary-Ring coefficients are supporting library extensions. These abstract statements use the existing transition maps. The different-label Rips relation consumer supplies the Lemma 4.3 face-image mechanism from an explicit absolute pairwise distortion bound, deriving dY≤dX+ε directly rather than relying on the first printed distance comparison without its additive allowance. It constructs the canonical interleaving over arbitrary rings and a full finite-diagram bound through the existing different-grid constructible theorem; these are supplied-data/library extensions, not a Gromov–Hausdorff selection theorem or whole real H₀ convention identification. RipsGromovHausdorff connects finite metric spaces to mathlib’s optimal GH coupling through a closed ambient Hausdorff correspondence of distortion at most twice the Hausdorff error. It realizes the finite specialization of §4.2/Lemma 4.3’s mechanism and proves full sublevel-diagram bounds dB ≤ 2 dGH, ambient finite sampling and convergence to a fixed finite target. The enumeration/order/universe bridges and finite-sample comparison through a common possibly infinite target are library consequences. RipsQTame formalizes the Rips part of Proposition 5.1: finite nets and the canonical interleaving law factor each strict structure map through finite-dimensional homology for every totally bounded metric space. The infinite compact unit interval has q-tame homology and infinite-dimensional H₀ at zero. Isometric relabelling gives natural module isomorphisms independently of vertex orders. RipsQTameDiagram consumes the separately proved integer-content extraction to obtain locally finite decorated multiplicities, including essential deaths and exact Rips rectangle/tail counts. This construction does not claim a whole-real H₀ convention equivalence; finite-diagram compatibility and Rips-specific q-tame metric bounds are recorded below. The general arbitrary-module setting retains the finite-birth scope described in the persistence-measure entry.
  • Additional module scope: RipsHausdorffInterleaving uses the §4.2/Lemma 4.3 correspondence mechanism for arbitrary isometric ambient images at finite Hausdorff distance, choosing near points with positive slack. It proves actual Rips homology interleavings above twice that distance and the resulting extended module-distance bound, over every ring and in every degree. Nonempty totally bounded images supply finiteness without closedness or completeness. The infinite open/closed interval fixture has zero module distance. The library retains its empty negative tail; the module interleaving alone does not assert the separately constructed decorated diagram or an infinite-space diagram bound. See Hausdorff module stability.
  • Rights/evidence: profile AR, v3 abstract license link/full-text banner; separate holder notice not established. Exact next step for text copying/AI input/reproduction: establish a licensed aligned manifestation or holder permission. This record supplies no universal reuse clearance.

CB14-v3 — real pfd interval decomposition (metadata inspected; primary comparison pending)

  • Identity/version: William Crawley-Boevey. Decomposition of pointwise finite-dimensional persistence modules. arXiv:1210.0819v3 metadata, revised 2014-07-29 (initial submission 2012-10-02; v2 2014-04-30). Related publication: Journal of Algebra and Its Applications 14(05), article 1550066 (2015), DOI 10.1142/S0219498815500668; Crossref metadata records online publication 2015-03-17 and print publication 2015-06. The key identifies the 2014 arXiv version, not the journal publication year; exact version-to-publication alignment is not established here.
  • Use/classification: The real interval decomposition note gives an original explanation from the Lean declarations and proofs of the real pfd interval-sum and finite-observation APIs. The attribution to Theorem 1.1 and the selected proof route, including compatible-lift Lemma 4.1, is retained from #506, not newly checked against the article text. The mathematical account is code/contract-based (digest-based); it does not claim a direct source match or full-proof comparison.
  • Evidence and rights: Bibliographic and rights metadata were inspected 2026-10-30; the article body was not consulted. The exact arXiv v3 page links the nonexclusive-distribution license (AR), which records a nonexclusive grant to arXiv and establishes no affirmative third-party article-text AI-input or reuse grant. The inspected Crossref response contains no license field; that absence supplies neither permission nor prohibition. Under #506, fresh article-text input or primary-proof review remains pending an aligned licensed manifestation or specific permission, plus exact version and hypothesis alignment. No article text or figures are reproduced here.

Sources for the next directions and historical comparisons

The following manifestations were inspected for bibliography, attribution and selected scope only. They are preparation/background sources; this register does not certify a new Lean theorem or a complete mathematical review of these papers.

FG10 — boundary transport

  • Identity: Alessio Figalli; Nicola Gigli. A new transportation distance between non-negative measures, with applications to gradients flows with Dirichlet boundary conditions (the consulted manuscript prints “gradients”). CVGMT record and manuscript; institutional publication record, DOI 10.1016/j.matpur.2009.11.005, J. Math. Pures Appl. 94, 107–130 (2010).
  • Manifestation/dates: CVGMT-hosted 28-page manuscript, unnumbered; its accepted/VoR status and upload date are not established. CVGMT record created 2009-11-10; this is evidence of a record date, not a verified earliest dissemination date. Publication year 2010 is independently recorded; precise first-public/publication dates remain unresolved.
  • Printed affiliations: Figalli: previous École Polytechnique, CMLS; manuscript-current University of Texas at Austin. Gigli: Université de Bordeaux (p. 1 footnotes). These are manuscript-era labels.
  • Use/classification: Prior boundary transport: Problem 1.1 uses a bounded open Euclidean domain with quadratic Euclidean cost; Lemma 2.1 and Theorem 2.2 give measure gluing and distance laws there. The library's finite occurrence matching laws are proved independently. This manuscript does not itself establish the later upper-half-plane, general-p or L∞ theory; no Lean measure-coupling result is attributed to it.
  • Rights/evidence: profile U. No explicit third-party reuse license or rights-holder notice was established in the inspected manuscript/record. Author hosting does not settle local copying, quotation, AI input or redistribution. Exact next step: clarify the manuscript's reuse terms with its rights holder, or identify an explicitly licensed equivalent manifestation; no equivalent licensed version is verified here.

OH17-v2 — linear learning and inverse analysis

  • Identity: Ippei Obayashi; Yasuaki Hiraoka. Persistence Diagrams with Linear Machine Learning Models. arXiv:1706.10082v2.
  • Manifestation/dates: v2, 2017-07-06; initial submission 2017-06-30. Full text contains placeholder journal/acceptance fields; those are not publication evidence. The related journal VoR, Persistence diagrams with linear machine learning models, has Ippei Obayashi; Yasuaki Hiraoka; Masao Kimura, in that order: Journal of Applied and Computational Topology 1(3–4), 421–449 (2018), DOI 10.1007/s41468-018-0013-5, online 2018-05-05, print 2018-06 (Crossref metadata). This three-author VoR is a separate manifestation; the consulted v2 remains the two-author preprint. “Obayashi–Hiraoka” attribution is not transferred to the VoR, and v2/VoR statement alignment remains unverified.
  • Printed affiliations: both Advanced Institute for Materials Research (WPI-AIMR), Tohoku University; Hiraoka also Center for Materials research by Information Integration (CMI2), MaDIS, National Institute for Materials Science (NIMS) (v2 author block). No Japanese name spelling or current affiliation is inferred.
  • Use/classification: #227 direction 2, attribution/background for combining PI with linear models and inverse analysis (§2.4–2.6, §3); cycle reconstruction and linear-learning applications remain separate future scope. No current Lean consumer.
  • Rights/evidence: profile AR, v2 abstract license link and full-text banner. Separate copyright-holder notice not established. The related VoR’s Crossref tdm/vor entries link Springer terms, not a CC license; the publisher preview identifies no CC grant. Exact next step for text copying/AI input: establish an explicitly licensed aligned manifestation or clarify holder permission; the identified VoR is not a verified licensed alternative.

MW16-v2 — categorical Mapper approximation

  • Identity: Elizabeth Munch; Bei Wang. Convergence between Categorical Representations of Reeb Space and Mapper. arXiv:1512.04108v2, full text.
  • Manifestation/dates: v2, 2016-04-12; initial submission 2015-12-13. Related SoCG 2016 article, DOI 10.4230/LIPIcs.SoCG.2016.53, published 2016-06-10, 53:1–53:16.
  • Printed affiliations: Munch: Mathematics and Statistics, University at Albany–SUNY; Wang: Scientific Computing and Imaging Institute, University of Utah (v2 p. 1).
  • Use/classification: Theorem 4.1, Lemma 7.1 and Corollary 8.1 supply the categorical Mapper colimit, resolution estimate and convergence contracts specialized to the real interval basis and finite graph-PL input. categoricalMapperData retains all cover-intersection components; its assembly colimit, interval-cosheaf and finite-constructibility proofs feed the existing finite reconstruction equivalence. The geometric comparison uses the graph-PL Reeb quotient and the maximum full width of retained bounded open intervals, with coverage only of the attained range. The same interval-basis argument is extended to constructible ℝ-spaces in universe zero, without Hausdorffness, with the actual Reeb quotient as endpoint through DMP15-v1 Theorem 3.17 and finiteness from the constructible path-component cosheaf; this is a library extension, not the full higher-dimensional source theorem. Ordinary Mapper identification and higher-dimensional Reeb-space approximation are outside this result. See categorical Mapper approximation.
  • Rights/evidence: v2 AR (full-text license banner). The separate SoCG manifestation is explicitly CC-BY 3.0 on its publisher landing page, with ©Elizabeth Munch and Bei Wang on p. 1. It is an available alternative requiring attribution, license identification and indication of changes; theorem/hypothesis/numbering alignment with v2 remains to be checked before substitution. Do not transfer its license to v2.

B15 — landscapes (background)

  • Identity: Peter Bubenik. Statistical Topological Data Analysis using Persistence Landscapes. JMLR 16(3), 77–102 (2015), article PDF.
  • Manifestation/dates: JMLR article, published 2015-01 (p. 1); initial arXiv:1207.6437 submission 2012-07-27. ArXiv v4 includes two examples absent from the journal article, per its version notice. No journal DOI established here.
  • Printed affiliation: Mathematics, Cleveland State University (article p. 1).
  • Use/classification: #227 direction 2 historical comparison: nonlinear order-statistic landscapes in a Banach space, not the PI additive surface/pixel theorem. No current Lean consumer or landscape implementation commitment.
  • Rights/evidence: ©2015 Peter Bubenik (p. 1). JMLR author policy, Publication states a public CC-by grant and author-retained copyright, but the inspected article does not specify a CC version. Profile U for version-specific reuse: clarify the applicable license version with JMLR/holder before claiming CC4 permissions or sending article text as AI input. No figures/code are reproduced.

RHBK15-v1 — scale-space kernel (background)

  • Identity: Jan Reininghaus; Stefan Huber; Ulrich Bauer; Roland Kwitt. A Stable Multi-Scale Kernel for Topological Machine Learning. arXiv:1412.6821v1, full text.
  • Manifestation/dates: v1/initial submission 2014-12-21. Related CVPR publication DOI 10.1109/CVPR.2015.7299106, pp. 4741–4748: IEEE-deposited Crossref metadata gives print 2015-06, while the ISTA institutional record gives Date Published 2015-10-14. These are distinct metadata fields, not a verified first-online date. Publisher retrieval, license and edition differences remain unresolved.
  • Printed affiliations: Reininghaus/Huber: IST Austria; Bauer: IST Austria, TU München; Kwitt: University of Salzburg, Austria (v1 p. 1; the PDF, unlike extracted HTML, clearly groups Reininghaus and Huber).
  • Use/classification: #227 direction 2 historical comparison, §4 signed reflected-Gaussian scale-space kernel and its stability. It is distinct from a nonnegative weighted PI surface and pixel vector. No current Lean consumer. No single overall invention claim is made: the PI article also acknowledges size-function surface predecessors; those papers are not independently used by this library and are not silently treated as inspected sources.
  • Rights/evidence: profile AR, v1 abstract license link/full-text banner; separate holder notice not established. Exact next step for text copying/AI input: inspect the published license and alignment, or clarify holder permission. No source code/data license is inferred from the paper.

Pinned software sources

Software projects have contributor/file notices rather than inferred academic affiliations; no project-wide author ordering or first-public date is claimed without citation metadata. Revision dates are not publication dates. Preserve the notices of any adapted file.

Lean4-4.33.1

Lean 4, v4.33.1, the revision consulted by the digests that cite this entry; it is no longer the installed pin (see Lean4-4.35.0). Lean project contributors, no academic affiliation inferred. LICENSE: Apache-2.0; individual notices and separately licensed bundled components require their own checks if redistributed. Release date/earliest public date not established here. Foundation and proof checker, not a paper-derived TDA theorem. AP2 for the licensed Lean code.

mathlib-0df444a

mathlib4, commit 0df444a360eaa60ab8c11dca51a86af692955474, the revision consulted by the digests that cite this entry; it is no longer the installed pin (see mathlib-56a0058). mathlib community contributors, file-level copyright attribution. Exact LICENSE, Apache-2.0, independently checked against the pinned checkout at the time. AP2; source-level consumers and declarations are mapped below. The signed contiguity prism adapts the finite-partition cancellation proof from SimplicialObject/ChainHomotopy.lean; its module preserves Fabian Odermatt's 2025 copyright and Apache-2.0 notice and identifies the changes. The topologist's sine-curve fixture adapts Counterexamples/TopologistsSineCurve.lean from the same revision; its module preserves Daniele Bolla's 2025 copyright, the authors (Bolla and Loeffler) and the Apache-2.0 notice, and identifies the changes. No academic affiliation/current institution or project initial-public/release date is inferred from the commit.

Lean4-4.35.0

Lean 4, v4.35.0, as pinned in lean-toolchain; Lean project contributors, no academic affiliation inferred. LICENSE: Apache-2.0; individual notices and separately licensed bundled components require their own checks if redistributed. Release date/earliest public date not established here. Foundation and proof checker, not a paper-derived TDA theorem. AP2 for the licensed Lean code.

mathlib-56a0058

mathlib4, commit 56a005855ea17e0b0958f73ec801ae140b0a1679, pinned installed dependency; mathlib community contributors, file-level copyright attribution. Exact LICENSE, Apache-2.0, independently checked against the local pinned checkout. AP2. No academic affiliation/current institution or project initial-public/release date is inferred from the commit. Digests written against this pin cite this entry directly; digests written against the prior pin keep citing mathlib-0df444a as the revision they consulted.

TauCeti-5283b03

TauCeti, inspected commit 5283b03c90ba08f975b65098ee412a890b567a6e, LICENSE Apache-2.0 (AP2). Project contributors/file notices; no author ordering, academic affiliation or initial-public/release date inferred. Simplicial-map source comparison only; not an installed dependency. Its toolchain is Lean 4.35.0-rc3, a release candidate distinct from the library's stable 4.35.0. No TauCeti code is imported by this citation.

TauCeti-7b0d106

TauCeti, inspected commit 7b0d106bf79dbea38ba8819985bd029989181287. The directly read SimplicialComplex/ChainComplex.lean, Maps.lean and Realization/CWComplex/OrderedMap.lean carry Apache-2.0 (AP2) contributor notices. The first gives ordered Ring-valued chains and boundary square-zero; the inspected map modules do not give arbitrary signed chain action, and the cell-image and enumeration results require local injectivity or order conditions. This scoped inventory motivates the independently proved signed face vertex-map bridge; no code or dependency is imported. Its toolchain is Lean 4.36.0-rc1, distinct from the library's pin. No academic affiliations, author ordering or release dates inferred.

The separately inspected Representation/Basic.lean defines QuiverRep as Paths Q ⥤ ModuleCat k. Its general path-algebra module equivalence and finite-dimensional Krull–Schmidt infrastructure are not dependencies of the finite-line and based-matrix adapters, which use mathlib's path category directly. This is an interface comparison, with no TauCeti proof copied or imported.

ReebOrdering-aa7d10f

uda-lab/reeb-ordering, commit aa7d10f4f7dc0474b76f17ea81afcab5a230d521; source file blob 6b79160f368a9d6fd63f2d52f1db4d9c5697cfa8. LICENSE: Copyright 2026 Tomoki Uda, Apache-2.0 (AP2), independently checked at that revision. No academic affiliation/current institution or first-public/release date inferred. GraphRealisation.lean is an adapted construction source, not an imported dependency; reeb-graph-realization.md, reeb-graph-cycle-obstruction.md and reeb-graph-tree-contraction.md record adapted declarations and changes, with attribution retained. The license grants code reuse under its notice, license and change requirements, not rights in other papers.

Existing declaration and proof mappings

These mappings retain theorem numbers, conventions and explicit library extensions from the mathematical digests. Paper citation keys resolve to the manifestation cards above; software links resolve to immutable versions. A source reference is not a claim that its entire article or all its algorithms have been formalized.

  • CB14-v3 — The real interval decomposition note explains the implemented PModule.{u,0,u} k ℝ scalar-sum and finite-observation results from their Lean declarations and proofs, with the attribution retained in #506 stated separately from that code/contract evidence. Primary-text comparison remains pending in #506; this is a digest-based account, not a new source match or full-proof certification.

  • BOOS24-v3 — Botnan–Oppermann–Oudot–Scoccola, On the bottleneck stability of rank decompositions of multi-parameter persistence modules, Proposition 5.1: TDA.Persistence.GeneratedUpsetResolution constructs the actual join-indexed resolution and proves ordinary projectivity, chain laws, and exactness at every lattice index via the simplex on active generator occurrences. GeneratedUpsetProjectiveDimension packages its canonical categorical projective resolution and proves the actual ordinary projective-dimension upper bound m−1. Repeated values retain their occurrence indices. Minimality and global-dimension results are not claimed; see the proof exposition.

  • BOOS24-v3 — Theorem 6.1, §6.1 and its proof, together with §§2.7–2.9 and Proposition 4.3: TDA.Persistence.RankSchanuelUnshifted formalizes the actual unshifted parity-sum interleaving step, using the existing shifted Schanuel comparison and the proved 2n−2 bound on zero-padded terms of every supplied minimal resolution. TDA.Persistence.RankSignedStability constructs actual finite hook presentations of those sums, identifies their occurrence multisets with the signed cross-sign sums, and proves actual finite matching and the squared (2n−1)^2 extended-distance bound under the original finite-presentation hypotheses. TDA.Persistence.RankSignedBarcodeOneDimensional consumes the actual degree-zero hook presentation and the existing finite-constructible isometry to prove the ordinary one-dimensional barcode realization and bottleneck/interleaving isometry under the original small finite-presentation hypothesis. TDA.Persistence.RankSignedNontriangle formalizes the concrete two-coordinate signed cross-sign triangle failure of BOOS v3 Remark 6.3, using the staircase in Oudot--Scoccola v4, Example 4.1 rescaled by three, and proves its realization by actual originally finitely presented modules. This does not identify the signed hook dissimilarity with ordinary bottleneck, prove the countable extension, establish optimality of (2n−1)², or attach a metric instance. See unshifted Schanuel interleavings, signed hook bottleneck stability, and the one-dimensional barcode bridge, and the signed nontriangle example.

  • BOOS24-v3 — §2.4, Definition 5.4, Notation 5.9, and Lemmas 5.6–5.7: TDA.Persistence.IncidenceQuotientBaseChange realizes the right-module quotient base-change construction for the actual interval-pair incidence algebra by factoring the action through its quotient. The library proves its quotient universal property, projective and finite-colimit preservation, and the canonical degree-zero vertex comparison. TDA.Persistence.IncidenceQuotientDerivedOne proves exactness in degree one after applying extension to any projective resolution of a restricted quotient module, and the resulting first left-derived vanishing. TDA.Persistence.IncidenceQuotientColumnAction realizes the higher-degree projective-ideal/column mechanism of Lemma 5.6 by an additive functor and a natural assembly with image the actual quotient relations and injective on projectives. TDA.Persistence.IncidenceQuotientDerived proves exactness at every degree n ≥ 2 after extension of any projective resolution of any incidence module, with actual categorical left-derived vanishing there; for restricted quotient modules it proves exactness and vanishing in every positive degree. These statements leave the unaugmented degree-zero comparison unchanged and make no numerical dimension or 2n−2 persistence claim. See right-module base change along the diagonal quotient. The minimal-complement count used in Theorem 5.16 is formalized separately: TDA.Persistence.FiniteLatticeComplementCardinality proves the product, adjoined-top, and actual interval-pair counts, while TDA.Persistence.FiniteGridInjectiveRankDimension derives the arbitrary-field ordinary-injective Hom bound from them. These generator-count results and the ordinary-injective Hom projective-dimension corollary do not alone establish the finite-dimensional actual incidence-module bound. TDA.Persistence.IncidenceVertexProjectiveDimension separately proves that bound and the actual 2n grid bound using a finite-free comparison through the established all-vertex representation and representation-category theorem; it does not establish a global-dimension bound for arbitrary incidence modules or the universal relative 2n−2 persistence conclusion. See minimal-complement cardinalities in finite grids and ordinary projective dimensions of actual incidence modules.

  • BOOS24-v3 — The ordinary representation-category dimension component around Corollary 5.3, Proposition 5.15 and Theorem 5.16 is realized by TDA.Persistence.OrdinaryFiniteLatticeDimension: for an arbitrary field and finite bounded lattice, it proves that every actual pointwise finite-dimensional persistence functor has ordinary projective dimension at most the maximum minimal-complement count. The library derives this through actual hook rows, ordinary injective bounds and a discharged finite uniform bound. It also consumes the proved grid counts to give the ordinary n and 2n representation bounds. This is not a theorem for arbitrary incidence-algebra modules and does not by itself supply Hom-module bounds or the source's broader conclusion. The actual finite-grid Hom-module/Hom-cokernel result is a separate specialization; see ordinary projective dimension on finite lattices and finite-grid Hom-module and Hom-cokernel dimensions.

  • DMW16-v2 — Dey–Mémoli–Wang, Multiscale Mapper: Topological Summarization via Codomain Covers, §2–3: indexed cover maps, pullbacks and refinement towers underlying the set-theoretic cover interface; the library interface permits arbitrary sets and support-aware range coverage. §4.1: interleaved cover towers give interleaved Mapper homology modules, in Mapper foundations, with bottleneck bounds for step towers on finite grids as a library specialisation in Mapper foundations.

  • CO16-v3 — Carrière–Oudot, Structure and Stability of the 1-Dimensional Mapper, §2.5 and §3: ordinary Mapper as the nerve of the path-component pullback cover and the distinct MultiNerve construction that retains path-component data for intersections (the paper writes "connected" under its stated path-connected convention).

  • LOT21-v4 — Leygonie–Oudot–Tillmann, A Framework for Differential Calculus on Persistence Barcodes, Lemma 4.10 and preceding definition (9): finite equality-frontier argument for a dense local-order-stability locus. Order chambers extract this topological argument for continuous finite-coordinate families; barcode differentiability is outside this result.

  • CSEH07 — Cohen-Steiner, Edelsbrunner and Harer, Stability of persistence diagrams, §2, pp.106–108: persistent-rank multiplicities and triangle counts at distinct noncritical thresholds. The finite simplicial sublevel construction, including critical and self thresholds under closed-birth/strict-death conventions, is described in finite-sublevel-diagram.md.

  • CCDS14-v1 — Chazal–Crawley-Boevey–de Silva, The observable structure of persistence modules, v1 §1.1, p.1: the inclusive finite interval [p,q] corresponds to the library's half-open index bar [p,q+1). Relabelling terminal deaths by ⊤ and proving the tail inclusion identity are finite-sublevel library results, described in finite-sublevel-diagram.md.

  • CCDS14-v1 — Chazal–Crawley-Boevey–de Silva, The observable structure of persistence modules, §1.1, p.1: closed discrete intervals, their interval modules, and the four-rank multiplicity formula. The finite half-open translation and supplied-barcode scope are summarized in finite-interval-rank.md. The decomposition-free nonnegativity argument for arbitrary finite-dimensional modules is summarized in finite-rank-nonnegativity.md; that proof is a library construction, not a theorem attributed to the paper. The finite telescoping recovery of map ranks is summarized in finite-rank-reconstruction.md; it is likewise a library construction.

  • mathlib-0df444a — Mathlib preabstract simplicial complexes: support-aware face predicate, lower-set property, empty bottom and the ambient-singleton distinction.

  • mathlib-0df444a — Mathlib free Finsupp basis: Finsupp.basisSingleOne.
  • mathlib-0df444a — Mathlib chain constructors: ChainComplex.of and ofHom.
  • mathlib-0df444a — Mathlib ModuleCat short-complex homology: moduleCatMk, moduleCatToCycles, moduleCatHomologyIso.
  • mathlib-0df444a — Mathlib matrix coordinates: LinearMap.toMatrix and its coordinate formula; composition.
  • mathlib-0df444a — Mathlib submodule map and comap, range, restricted equivalences, and matrix function linear maps: the finite-cycle and boundary transport described in simplicial-finite-cycles.md.
  • mathlib-0df444a — Mathlib rank-nullity, submodule map and comap, and finrank monotonicity: pinned linear-algebra ingredients for the restricted-image proof in finite-rank-nonnegativity.md.
  • mathlib-0df444a — Mathlib canonical submodule quotient maps and equivalences: Submodule.mapQ and Submodule.Quotient.equiv descend the cycle-coordinate equivalence and matrix-induced inclusion through boundaries, as described in simplicial-finite-homology.md.
  • mathlib-0df444a — Mathlib ring-homomorphism factorization through a surjection, ModuleCat change of rings, projective objects under adjunction and left-derived functors: RingHom.liftOfSurjective, Submodule.mapQ/liftQ, scalar restriction, the epimorphism criterion for projective preservation and leftDerivedZeroIsoSelf support the actual quotient construction and degree-zero comparison in right-module base change along the diagonal quotient. The general commutative-only ModuleCat.extendScalars API is not used for the incidence rings.
  • Finite coordinate bridge for simplicial chains: degreewise signed boundary and same-ambient inclusion matrices, their coordinate squares, finite theory assumptions, and face fixtures.
  • mathlib-0df444a — Mathlib simplicial homotopy cancellation: finite-index partition proof adapted in TDA.Simplicial.ContiguityHomotopy, with the upstream notice preserved; homotopy and short-complex homology: homotopy-to-homology equality used by the signed contiguity bridge.
  • mathlib-0df444a — Mathlib SSet chain complex: chainComplexFunctor, chainComplex, homology maps and functor.
  • mathlib-0df444a — Mathlib ModuleCat colimits: coproduct support needed for the SSet route.
  • mathlib-0df444a — Mathlib normalized Moore complex, alternating-face complex and natural inclusion, and Dold–Kan homotopy equivalence: abstract normalization, not a finite face-basis identification.
  • TauCeti-5283b03 — TauCeti simplicial maps at inspected revision and toolchain: source reference only; the recorded TauCeti toolchain is Lean 4.35.0-rc3.
  • Bauer21-v2 — Bauer, Ripser: efficient computation of Vietoris–Rips persistence barcodes, v2 §3.1: maximum-nonzero-row pivot convention and column reduction.
  • ZC04 — Zomorodian and Carlsson, Computing Persistent Homology, §4: field-based column operations in the persistence computation.
  • mathlib-0df444a — Mathlib matrix transvections: pinned right-multiplication identities for the target column and unchanged non-target columns.
  • DMP15-v1 — de Silva, Munch and Patel, Categorified Reeb Graphs (arXiv:1501.04147v1, HTML v1): §2.2, constructible ℝ-spaces as quotients of vertex pieces and cylinders 𝔼ᵢ × [aᵢ, aᵢ₊₁] by the two endpoint identifications, used here with arbitrary topological pieces, compactness and local path-connectedness as hypotheses, and Hausdorffness proved from Hausdorff pieces, summarized in reeb-constructible-space.md, Remark 2.3, refinement of the critical set by splitting cylinders, proved for finite data with every vertex set, edge set and incidence of the refined data identified, summarized in reeb-rgraph.md, and Lemma 2.4, the cylinder principle, proved for arbitrary pieces by a clamp homotopy as ordinary homotopy equivalences whose forward maps are the fiber inclusions, with the incidence diagrams on path components and without the zero-based edge indexing of its parenthesis, summarized in reeb-constructible-cylinder.md; §2.3, ℝ-graphs as constructible ℝ-spaces with finite discrete pieces given by finite vertex and edge sets with attaching maps, refined by splitting edges at new vertices (Notation 2.7), the full subcategory Reeb of ℝ-Top, and Proposition 2.8, morphisms over a common critical set correspond bijectively to consistent vertex and edge maps, with the union of critical sets as a common refinement and the resulting description independent of the presentations, summarized in reeb-rgraph.md; the quotient of vertex points and edge intervals of Notation 2.7 is also the finite, height-oriented context of the realization of arbitrary simple graphs by dart cells, whose construction is adapted from reeb-ordering and whose path-component equivalence and naturality are new here; none is attributed to the paper, summarized in reeb-graph-realization.md; §2.4, ℝ-spaces, the Reeb quotient by path components of levelsets, the quotient principle (Lemma 2.9), the Reeb functor (Proposition 2.10) and its point, a natural transformation from the identity making it a pointed endofunctor (Observation 2.11), Proposition 2.12, the Reeb functor carries constructible ℝ-spaces to ℝ-graphs, proved with the path components of the pieces as finite data, the levelset-preserving bijection from the Reeb quotient to their graph and the compact-to-Hausdorff theorem, from compact locally path-connected pieces without Hausdorff pieces or conditions on the endpoint maps, together with the full subcategory ℝ-Topᶜ of §2.2 with all morphisms of ℝ-Top and the Reeb functor read after Proposition 2.13 as a functor ℝ-Topᶜ → Reeb, summarized in reeb-constructible-quotient.md, and Proposition 2.13, every ℝ-graph is naturally isomorphic to its Reeb graph, summarized in reeb-rgraph.md; §3 opening, the category Int of open intervals with inclusions; §3.1 Example 3.3 and §3.4, path components of inverse images over open intervals, maps induced by inclusions, and natural transformations induced by height-preserving maps; Notation 2.7 and Example 3.3 are also the context of the identification, for the piecewise-linear height of the realization of an arbitrary simple graph and every open interval, of path components of the preimage with components of the incidence graph of surviving vertices and active edges, natural in the interval, with a vertex-only form under endpoint coverage; it is a library construction and is attributed to none of Proposition 3.12, Theorem 3.17 and Proposition 3.18 of the paper, summarized in reeb-graph-pl.md; §3.2, the cosheaf gluing condition, diagram (3.4), and its universal-property reading (3.5)–(3.6); §3.3 Definition 3.9, constructible cosheaves, used with the nonempty-source correction, Notation 3.11, the vertex and edge values on stars and gaps with their endpoint maps, Proposition 3.12, natural transformations between constructible cosheaves correspond bijectively to commuting short-interval maps, Remark 3.13, the common critical set, and Corollary 3.14, the short-interval isomorphism test, summarized in reeb-constructible-cosheaf.md; §3.4 Proposition 3.15, gluing of the path-component pre-cosheaf, proved by local constancy along paths; the library proves the path step of that proof by finite subdivision and the gluing itself in factored form; Proposition 3.16, the path-component cosheaf of a constructible ℝ-space is constructible with the same critical set, proved from finitely many path components of every piece, summarized in reeb-constructible-cylinder.md; Theorem 3.17, the functors 𝒞' and 𝒞''ℛ from ℝ-Topᶜ to Cshᶜ are naturally isomorphic, with the comparison η = 𝒞ρ of its proof obtained by applying the path-component functor to the canonical projection, proved for η restricted to open intervals as a natural isomorphism into Cshᶜ on all morphisms of ℝ-Topᶜ, and for presentations whose pieces are locally path-connected with finitely many path components, without compactness, by the short-interval test applied to the map to the component graph, summarized in reeb-constructible-comparison.md; Proposition 3.18, a constructible cosheaf realizing finite incidence data, with labelled uniqueness for arbitrary data obtained together with the map classification of Proposition 3.12, and the reconstruction of every constructible cosheaf from its extracted data, summarized in reeb-constructible-representation.md; Example 3.20, where gluing fails for connected components; Example 3.21, the topologist's sine curve, where 𝒞(𝕊, x) and 𝒞ℛ(𝕊, x) are not isomorphic; §3.5 Theorem 3.22, the restriction 𝒞'' of the path-component cosheaf functor to Reeb is an equivalence onto Cshᶜ, with full faithfulness over a common critical set through Propositions 2.8 and 3.12 and the cylinder diagrams, and essential surjectivity through the data of a constructible cosheaf and Corollary 3.14, stated at value universe 0, and the display locale functor 𝒟 of the paragraph after its proof, defined from the same data at a critical set with 𝒞𝒟 ≅ 1 and unique up to canonical natural isomorphism, summarized in reeb-rgraph-cosheaf-equivalence.md; §4.1–4.2, the expansion I^ε = (a - ε, b + ε) for ε ≥ 0 and the expansion functor Ω_ε with its unit; Definition 4.1 and Observations 4.10 and 4.12 (with the unit of Observation 4.11), ε-interleavings of pre-cosheaves and their form as natural transformations; Definition 4.4, the interleaving distance as the infimum of the interleaving parameters, with the empty infimum ∞; Proposition 4.5, the interleaving distance is an extended pseudometric, by composition of interleavings, whose displayed reverse map ψ³_I = ϕ¹_(I^ε₂) ∘ ψ²_I should read ψ¹_(I^ε₂); Theorem 4.6(i), stability of the component cosheaves under ‖f − g‖_∞ ≤ ε, proved by the inclusions f⁻¹(I) ⊆ g⁻¹(I^ε) and the maps they induce on path components, which bound the interleaving distance of the component cosheaves by ‖f − g‖_∞; Definition 4.14, Observation 4.15 and Propositions 4.16–4.17, the smoothing functor, its unit, its composition law and its nonexpansiveness; Theorem 4.18, smoothing restricts to endofunctors of Csh and Cshᶜ, split into Proposition 4.20 below and Proposition 4.19, smoothing carries constructible pre-cosheaves to constructible pre-cosheaves, with the critical set (S − ε) ∪ (S + ε) of its proof; Proposition 4.19 is false under the literal Definition 3.9, which allows an empty source interval, since the point cosheaf satisfies that definition at {0} while its 1-smoothing satisfies it at no finite critical set; Proposition 4.19 and the Cshᶜ half of Theorem 4.18 are proved for the corrected nonempty-source predicate, at the sorted shifted levels and for every value universe, with the enlargement maps and composition and commutation isomorphisms lifted to Cshᶜ as library statements, summarized in reeb-constructible-smoothing.md; Proposition 4.20, smoothing carries cosheaves to cosheaves, proved, for two disjoint cover intervals whose thickenings meet, by a finite chain of cover intervals across the gap whose thickenings contain the common overlap, and Remark 4.21, the argument is not specific to sets; §4.3 Definition 4.22, the thickening functor T_ε with X_ε = X × [-ε, ε] and f_ε(x, t) = f(x) + t, whose displayed T_ε[α] : (X_ε, f_ε) → (Y_ε, f_ε) should read (Y_ε, g_ε), Observation 4.23, the zero section τ^ε, Theorem 4.24, 𝒞T_ε ≅ S_ε𝒞 with the comparison (ρ_f)_I = π₀[p_I] of its proof, and Lemma 4.25, the first projection restricts to homotopy equivalences f_ε⁻¹(I) → f⁻¹(I^ε), proved with a clamp for λ for every open interval, including ∅, rays and ℝ, and every ε ≥ 0, with the comparison bijective, natural in intervals and in height-preserving maps, and, preceded by π₀ of the zero section, equal to the smoothing unit, which is the projection form of the square in the proof of Proposition 4.35, together with the reparametrization {(x, u) | u − ε ≤ f(x) ≤ u + ε} of the proof of Theorem 4.27 (Figure 8), summarized in reeb-geometric-thickening.md; Definition 4.22 is also the context of the library's addition map T_ε T_δ → T_(ε+δ), with its lax associativity, unit and zero-section laws, its π₀ compatibility with the composition isomorphism of Proposition 4.16, and the one-point non-isomorphism T_1 T_1 P ≇ T_(1+1) P; no source statement is claimed by it, and the semigroup isomorphisms of Observation 4.33 are proved independently only after the Reeb quotient, in reeb-quotient-thickening-absorption.md, summarized in reeb-geometric-thickening-addition.md; Proposition 4.26, the thickening of an ℝ-graph is a constructible ℝ-space, proved for every graph object and every ε ≥ 0, at ε = 0 through T_0 ≅ 1 and for ε > 0 by the sliding-window construction of the proof of Theorem 4.27 applied to graph input, instead of the source's piecewise linear argument, with the closed windows f⁻¹[u − ε, u + ε] presented exactly as quotients of finite sums of points and closed intervals, the shifted levels (S − ε) ∪ (S + ε) as critical set, affine stretches of edge fragments as cylinder maps, and the compact-to-Hausdorff step of Step 6 with Hausdorffness from the graph input; Step 4 of that proof writes F_k = f⁻¹(b) for some b ∈ (b_k, b_{k+1}), where its displayed decomposition and the maps of Step 5 require the closed window f⁻¹[b − ε, b + ε], and the endpoint attaching maps of Step 5 need not be injective; Theorem 4.27 beyond graph input is not part of that graph route, summarized in reeb-rgraph-thickening.md; Theorem 4.27, the thickening of a constructible ℝ-space is constructible, proved in general form for every constructible ℝ-space and every ε ≥ 0, with pieces that need not be Hausdorff, by a thickening presentation whose pieces are closed windows carrying the quotient topology of their raw pieces and a homeomorphism whose inverse is continuous by descent along the quotient map of the thickened raw pieces, in place of the compact-to-Hausdorff step of Step 6, with Proposition 4.26 recovered as a specialization; Corollary 4.28, that f_ε has critical set (S − ε) ∪ (S + ε), proved in the sense that the thickening is presented over these values in increasing order as an admissible critical set, without minimality or essentiality of every value; the point-set tools, presentation morphisms on a common grid, one-cylinder and general thickening presentations used for it are library statements, with no further source statement claimed by them, summarized in reeb-thickening-constructible.md; Theorem 4.24 restricted to constructible ℝ-spaces, 𝒞'T_ε ≅ S_ε𝒞' as functors ℝ-Topᶜ → Cshᶜ, natural in all height-preserving maps and lifted from the comparison ρ without a new geometric argument, with the zero section, ε = 0 and graph specializations, and, distinct from it, the composite of the inverse of the comparison of Theorem 3.17 at the thickening with that restriction, 𝒞''ℛT_ε ≅ S_ε𝒞', a library statement extending the proof of Proposition 4.32 to constructible input that restricts on graph objects to the already-proved Proposition 4.32, with the unit of Observation 4.34 in the form of Proposition 4.35, summarized in reeb-thickening-cosheaf-comparison.md; §4.4 Definition 4.30, the Reeb smoothing functor U_ε = ℛT_ε, defined on graph objects as the thickening, constructible by Proposition 4.26, followed by the Reeb functor; Example 4.31, items (1) and (3), the line stretched by ε in both directions and the loop of height at most 2ε whose thickening has one component in every levelset, proved for every interval value and for the height fibers; Proposition 4.32, 𝒞''U_ε ≅ S_ε𝒞'', with the comparison of its proof, the inverse of the comparison of Theorem 3.17 at the thickening followed by the comparison of Theorem 4.24, natural in every graph map, with its hom-component equation and the representative formula [q(x, t)] ↦ [x]; Observation 4.33, the contraction d_R(U_εf, U_εg) ≤ d_R(f, g) for the interleaving distance of the component cosheaves, and U_0 ≃ 1, constructed from T_0 ≅ 1 and the Reeb projection of graph objects, without its semigroup isomorphisms; Observation 4.34, the unit ζ^ε as the zero section followed by the Reeb quotient; and Proposition 4.35, whose bottom arrow is drawn as the inverse of the comparison of Proposition 4.32, stated in the equivalent form 𝒞(ζ) followed by the comparison equals σ and proved by exact cancellation of the comparison of Theorem 3.17 instead of the source's homotopy; the identification of the fibers and vertices of U_εf after Definition 4.30, which the source derives from Theorem 4.27 and Corollary 4.28, is not claimed in general, summarized in reeb-rgraph-smoothing.md; Definition 4.36, of which only the statement shape is attributed, proved here independently on graph objects together with the distance as an infimum in the shape of Definition 4.4 and the library inequality from the cosheaf interleaving distance to the geometric one, summarized in reeb-geometric-interleaving.md, and with the library equality of the geometric and cosheaf interleaving distances, composition of geometric interleavings and the triangle inequality, as library statements; the enlargement U_ε ⟶ U_δ, monotonicity, self-smoothing d_geo(A, U_ε A) ≤ ε and nonexpansiveness of U_ε for d_geo are library statements without a numbered source, summarized in reeb-geometric-interleaving-equivalence.md; Definition 4.30 and Observation 4.33 are also the context of the library's quotient comparison of the addition map, the homeomorphism ℛT_εT_δ ≅ ℛT_(ε+δ) with forward map ℛ of the addition map, its coherence, and the natural isomorphism from ℛT_εT_δ to U_(ε+δ) on graph objects; no source statement is claimed by it, summarized in reeb-quotient-thickening-addition.md; Observation 4.33, first bullet, the semigroup isomorphisms, of which only the statement shape U_εU_δ ≅ U_(ε+δ) on graph objects is attributed and which is proved here independently through the library's absorption isomorphism ℛT_εℛ ≅ ℛT_ε at thickenings of graph objects and the isomorphism from ℛT_εT_δ to U_(ε+δ), with associativity, unit laws and compatibility with the composition isomorphism of Proposition 4.16 as library statements, summarized in reeb-quotient-thickening-absorption.md; for Example 3.21, non-injectivity of the comparison at ℝ is proved, non-isomorphism of the two cosheaves is not claimed, summarized in reeb-quotient-comparison.md.
  • mathlib-0df444a — Mathlib path components: ZerothHomotopy, ZerothHomotopy.lift, ZerothHomotopy.sound and pathConnectedSpace_iff_zerothHomotopy; paths: Path.map; open sets: TopologicalSpace.Opens.
  • mathlib-0df444a — Mathlib path-connectedness: JoinedIn, joinedIn_iff_joined, pathSetoid and ZerothHomotopy; quotient homeomorphisms: Homeomorph.Quotient.congrRight and Homeomorph.quotientBot.
  • mathlib-0df444a — Mathlib topologist's sine curve (outside the Mathlib library target): the definitions S, Z, T, w and the non-path-connectedness proof, adapted in TDA.Reeb.QuotientComparisonSineCurveExamples with the graph restricted to 0 < x ≤ 1, as described in reeb-quotient-comparison.md.
  • mathlib-0df444a — Mathlib simple-graph components: SimpleGraph.ConnectedComponent, connectedComponentMk and ConnectedComponent.map, with nonemptiness part of SimpleGraph.Connected; graph homomorphisms on darts: SimpleGraph.Hom.mapDart. Used by the realization of simple graphs in reeb-graph-realization.md. Unoriented edges of darts: SimpleGraph.Dart.edge, Dart.edge_mem, Dart.edge_symm and dart_edge_eq_iff; with SimpleGraph.induce, ConnectedComponent.lift, SimpleGraph.IsBipartiteWith and restricted quotient maps, Topology.IsQuotientMap.restrictPreimage_isOpen, used by the incidence-graph comparison in reeb-graph-pl.md.
  • ReebOrdering-aa7d10f — uda-lab/reeb-ordering, ReebOrdering/GraphRealisation.lean at commit aa7d10f4f7dc0474b76f17ea81afcab5a230d521 (blob 6b79160f368a9d6fd63f2d52f1db4d9c5697cfa8), Copyright 2026 Tomoki Uda, Apache License, Version 2.0: construction source of the dart-cell realization of simple graphs, its lift and its edge and walk paths, adapted and not imported; the adapted declarations and changes are listed in reeb-graph-realization.md.
  • ReebOrdering-aa7d10f — uda-lab/reeb-ordering, ReebOrdering/GraphRealisation.lean at the same commit and blob: source of the one-edge winding map, its vanishing on deleted-edge walks and the covering-lift proof that a simply connected realization forces acyclicity, adapted and not imported; the adapted declarations and changes are listed in reeb-graph-cycle-obstruction.md.
  • ReebOrdering-aa7d10f — uda-lab/reeb-ordering, ReebOrdering/GraphRealisation.lean at the same commit and blob: source of the leaf collapse, the leaf slide, the leaf homotopy equivalence and the induction showing that the realization of a finite tree is contractible, with the finite tree criterion, adapted and not imported; the adapted declarations and changes are listed in reeb-graph-tree-contraction.md. The source's Hasse-order result is not ported.
  • mathlib-0df444a — Mathlib order-connected sets: Set.OrdConnected; convexity on ℝ: Convex.add, convex_iff_ordConnected; interval sums: Set.Icc_add_Icc; thickenings: Metric.thickening.
  • mathlib-0df444a — Mathlib unit-interval subdivision: exists_monotone_Icc_subset_open_cover_unitInterval, the positive-mesh lemma via lebesgue_number_lemma_of_metric; subpaths: Path.subpath, Path.range_subpath_of_le, Path.subpath_self and Path.concat; relations: Relation.ReflTransGen.
  • mathlib-0df444a — Mathlib over categories: Over, Over.mk, Over.homMk and Over.w; topological spaces as a category: TopCat.of and TopCat.ofHom.
  • mathlib-0df444a — Mathlib pairwise diagrams: Pairwise, Pairwise.diagram and Pairwise.cocone; colimits in Type: Types.isColimit_iff_coconeTypesIsColimit; colimit types: Functor.ColimitType and ιColimitType_eq_iff; pairwise sheaf condition: TopCat.Presheaf.IsSheafPairwiseIntersections.
  • mathlib-0df444a — Mathlib uniform extended distance: the extended supremum distance on α →ᵤ β for an arbitrary domain, UniformFun.edist_def, UniformFun.edist_le and the PseudoEMetricSpace (α →ᵤ β) instance; finite-indexed Pi distances, which do not apply to an arbitrary domain.
  • mathlib-0df444a — Mathlib reproducing kernel Hilbert spaces: RKHS, RKHS.kerFun, RKHS.posSemidef_tfae, RKHS.OfKernel and Moore's theorem RKHS.OfKernel.kernel_ofKernel (mathlib cites Paulsen–Raghupathi); Schur product theorem: Matrix.PosSemidef.hadamard over arbitrary index types. Used by weighted-kernel-embedding.md.

  • LOT21-v4 — Leygonie–Oudot–Tillmann, A Framework for Differential Calculus on Persistence Barcodes, Lemma 4.6 and Proposition 4.5: inverse-image restriction of a supplied interval decomposition under a continuous increasing real map and its use for ordering-equivalent barcode templates. The finite cut and decomposition-free rank transport construction is summarized in finite-rank-reindex.md; it is a library construction, not a real-index decomposition or template theorem attributed to the paper.

  • diagram transport at attained grades: the two occurrence-multiset laws for ordering-equivalent genuine finite-face filters and strictness only on attained values are direct constructions from the existing library definitions. They reuse mathlib increasing enumeration uniqueness and the existing canonical rank-barcode uniqueness theorem; no arbitrary ambient-threshold naturality or decomposition theorem is attributed to a paper.

SFL11 — C₀ universality and measure separation

  • Identity/version: Bharath K. Sriperumbudur; Kenji Fukumizu; Gert R. G. Lanckriet. Universality, Characteristic Kernels and RKHS Embedding of Measures. JMLR 12(70), 2389–2410 (2011), journal PDF. Submitted 2010-03, revised 2010-10, published 2011-07 (p. 1).
  • Printed affiliations: Sriperumbudur: Gatsby Computational Neuroscience Unit, University College London; Fukumizu: The Institute of Statistical Mathematics; Lanckriet: Department of Electrical and Computer Engineering, University of California, San Diego.
  • Use/classification: Proposition 2 relates C₀ density and finite signed-measure injectivity. GaussianMeasureEmbedding and GaussianUniversality specialize the endpoint to the full-plane Gaussian and the existing RKHS. The spectral proof uses mathlib's Gaussian characteristic function; norm_signedEmbedding_sq gives the Gaussian energy identity relevant to Proposition 4. The general kernel equivalences and the full radial-kernel equivalences of Proposition 5 are not formalized. Exact statements and weighted persistence-measure consumers.
  • Rights/evidence: U, inspected 2026-10-20. The PDF prints ©2011 the three authors; the inspected article page and PDF establish no article-specific third-party reuse grant. This entry supplies metadata, links and an original mathematical summary. No article text or figures are reproduced; broader copying, article input or redistribution requires the clarification described in U.

TauCeti-b971ef7 — signed C_c functional decomposition

  • Identity/version: The Tau Ceti contributors, TauCeti at b971ef7017efd17b28899a41289f2f11cdced2a2, specifically signed Riesz representation and the three toReal helpers in compactly supported continuous maps.
  • Use/classification: TDA.Topology.C0MeasureRepresentation adapts the necessary C_c positive/negative decomposition privately to the pinned mathlib, retaining the copyright, authors, license and change notice. Its public bounded C₀ representation adds finite-mass bounds and uniform cutoff density. The source snapshot pins mathlib e3e75e86143b43849829fdcd44a39a6b0fe35234; the adaptation was made against mathlib 0df444a360eaa60ab8c11dca51a86af692955474, tda-lib's pin at the time (now 56a005855ea17e0b0958f73ec801ae140b0a1679). The adapted source is not a package dependency. Compact-domain kernel universality in that snapshot does not supply the noncompact C₀ theorem.
  • Rights/evidence: AP2, inspected 2026-10-20, the pinned repository LICENSE and source headers. The inspected tree has no NOTICE file. The adapted file preserves the source notice and identifies its immutable source and modifications; tda-lib also distributes Apache-2.0 in its root LICENSE.

CDGO13-v3 — persistence rectangle measures

  • Identity/version: Frédéric Chazal; Vin de Silva; Marc Glisse; Steve Oudot. The structure and stability of persistence modules. arXiv:1207.3674v3, 20 March 2013; PDF.
  • Use/classification: PersistenceMeasure constructs the image-and-kernel quotient of Proposition 2.5 (§2.2), proves finite-dimensionality from q-tameness, the Proposition 2.3 four-rank identity, and coordinate-splitting additivity. PersistenceMeasureEssential constructs the essential birth-strip measure by eventual finite-target image quotient dimensions, following the boundary considered in §2.6. PersistenceMeasureBox proves the finite-rectangle inequality of Lemma 4.22 (§4.6) from actual interleaving maps, requiring only a finite expanded crossing image in its local form. This is an ingredient of Theorems 4.20–4.21; the finite-birth matching conclusion is formalized below through finite saturation and global compactness. See persistence rectangle measures and interleavings and rectangle measures.

PersistenceMeasureEssentialBox proves the positive-death boundary case of CDGO13-v3, Proposition 4.23 (§4.6), for q-tame source and target, directly from interleaving maps on image quotients. The essential value is then the attained minimum of §2.6. Its raw inequality of defined natural-valued Module.finrank minima needs only target q-tameness; that weaker assumption does not identify the source minimum with its genuine essential boundary count. No essential matching or diagram decomposition is assumed. See interleavings and essential birth strips.

IntegerContentAtoms formalizes the finite local atomic representation behind CDGO13-v3, Theorem 2.8, using compact open decorated cells. DecoratedMeasureCuts, DecoratedRectangleCells and DecoratedRectangleTopology retain boundary decorations. PersistenceMeasureBoxContent and PersistenceMeasureContent construct the actual module content from quotient split identities. PersistenceMeasureDiagram extracts finite-death multiplicities with exact rectangle counts; PersistenceMeasureEssentialContent, PersistenceMeasureEssentialDiagram and PersistenceMeasureExtendedDiagram add the essential boundary, exact tail counts and local finiteness. PersistenceMeasureDiagramCountability proves countable support, and PersistenceMeasureLowerBound transfers a module zero tail to a birth lower bound. LocallyFiniteMatchingCompactness proves algebraic bottleneck stability for the extracted finite-birth diagrams, and RipsQTameBottleneck identifies the finite-space Rips bottleneck with the bottleneck of the existing full finite diagrams. RipsQTameInfiniteMetricStability proves the exact factor-two ambient Hausdorff bound for complete q-tame diagrams of nonempty totally bounded Rips spaces, its compact Gromov–Hausdorff consequence, and convergence from finite samples of variable cardinality to a totally bounded target diagram. Isometric relabelling preserves the actual decorated diagram through its rectangle and essential-strip contents, without an order-preserving isometry assumption. These are Rips consumers of finite-net q-tameness and the algebraic stability theorem, consistent with the general metric stability results in §5.2; they retain the library's closed-diameter parameter and empty-negative-tail conventions. Birth-at-minus-infinity strata for unrestricted q-tame modules and finite-diagram compatibility for arbitrary q-tame modules remain open. See integer contents on decorated cells.

  • Rights/evidence: U, inspected 2026-10-23. The source is used for mathematical statements and an original formal proof; no source prose, figures or source code are redistributed.

PersistenceMeasureMatching formalizes the finite local matching step of CDGO13-v3, Theorem 4.25, Part 2(ii), §4.7, using occurrence counts and Hall's theorem. The strict source separation and lifetime hypotheses remain explicit; Example 4.26 prevents promoting a single pair of box bounds to general stability. See local matching from finite box counts.

BL24 — normalized subdivision and measure bifiltrations

  • Identity/version: Andrew J. Blumberg and Michael Lesnick, Stability of 2-Parameter Persistent Homology, Foundations of Computational Mathematics 24 (2024), arXiv:2010.09628.
  • Use/classification: Definitions 2.10–2.13 define normalized subdivision flags; Definitions 2.32–2.33 give the restricted interleaving category. MeasureBifiltration proves the finite-measure inclusion argument of Theorem 3.1 using mathlib's Lévy–Prokhorov inequalities. SubdivisionRips proves the finite sup-norm comparison step in Theorem 1.6(iii) and connects the concrete flag filtration to simplicial homology. FiniteGromovProkhorov proves the common-ambient finite-uniform GP witness reduction and arbitrary-outlier mass estimate. MulticoverCover proves the exact cover, its naturality and contractible finite intersections; MulticoverNerve proves that the natural homology comparison morphism is an isomorphism (cechHomologyToLargeFaceNerve_isIso), and GromovProkhorovMulticover proves the finite-uniform GP-to-SR homology interleaving bound (subdivisionRipsHomologyInterleavingDistance_le_gromovProkhorovEDist). Theorem 3.3's space-level equivalence remains unproved. Exact scope.
  • Rights/evidence: U, primary PDF inspected 2026-10-23. Original definitions, proofs, and summary are supplied; no source prose or figures are reproduced.

Sheehy13-v2 — sparse Rips weights, persistence and packing

  • Identity/version: Donald R. Sheehy. Linear-Size Approximations to the Vietoris–Rips Filtration. arXiv:1203.6786v2, revised 2013-03-06, versioned PDF.
  • Use/classification: TDA.Filtration.SparseRips uses the weight convention of §4, the deletion and contiguity arguments of §7, the finite tail-union mechanism of Lemma 8.1, the persistence comparison of Theorem 8.2, and the approximation and packing arguments of §9. The finite farthest-first permutation realizes the net geometry, with insertion radius ρ giving deletion time ρ/[η(1−2η)]. The natural sparse-to-relaxed homology equivalence and the logarithmic log((1−2η)⁻¹) interleaving give a persistence approximation in every degree. Continuity of the piecewise weights gives attained relaxed pair entries. A nonempty candidate face belongs to the sparse carrier exactly when its maximum pair entry is at most every finite vertex-deletion time, including ties. The resulting finite filter recovers the sparse filtration exactly. In positive degrees, finite-constructible algebraic stability bounds the log-diagram bottleneck by the same logarithmic shift. Separated-set packing bounds the full all-scales fixed-skeleton size. The coefficient-ring homology comparison uses the library's chain-homotopy bridge; the source states it over fields. A concrete rational H₁ cycle survives in the sparse complex after a pendant vertex has passed its deletion time, and its sparse-to-relaxed image is nonzero.
  • Rights/evidence: U for third-party reuse; the linked manifestation is used for mathematical reading and an original summary. No article text, figure or source code is reproduced here, and no broader article-reuse permission is claimed.

TauCeti-575dbdf — fixed-marginal plan compactness

  • Identity/version: The Tau Ceti contributors, OptimalTransport/Compactness.lean at 575dbdfaf4ae0df6e5bb35bb39e16458f028c91f.
  • Rights/evidence: AP2, pinned LICENSE and source header. The adaptation retains the source notice and identifies its revision and modifications; tda-lib distributes Apache-2.0 in its root LICENSE.
  • Use/classification: BoundaryTransportAttainment specializes the closed/tight probability-plan argument to the already-proved collapsed space and the pinned mathlib map interface. Native continuous-cost truncations, normalization and reverse comparison then prove finite-input attainment. No general coupling, cost or Wasserstein package is imported. The dependency map separates this source adaptation from incompatible ordinary-OT APIs and unresolved article-body use.

Fasy14 — confidence from finite-cover miss probabilities

  • Identity/version: Brittany Terese Fasy, Fabrizio Lecci, Alessandro Rinaldo, Larry Wasserman, Sivaraman Balakrishnan and Aarti Singh, Confidence sets for persistence diagrams, Annals of Statistics 42(6) (2014), 2301–2339, arXiv:1303.7117v3, DOI.
  • Use/classification: Lemma 4 and its proof motivate the finite-cover iid union bound in SamplingCoverage and its actual q-tame Rips consumer in RipsSamplingConfidence. The library retains the cover cardinality and minimum ball mass explicitly; it does not formalize the source's manifold covering-number or density estimates. Diameter-scale Rips stability gives factor two, whereas the source's distance-function/offset diagrams use factor one. See finite-sample Rips confidence.
  • Rights/evidence: U for third-party reuse. The primary statement and proof were read for an original formal proof and summary; no article text, figures or source code are redistributed.

Matching-distance affine arrangements

  • Sources: Bapat et al., Computing the Matching Distance of 2-Parameter Persistence Modules from Critical Values, arXiv:2210.12868v3, Lemma 4.2 and Theorems 3.1/4.4; Kerber–Lesnick–Oudot, Exact Computation of the Matching Distance on 2-Parameter Persistence Modules, arXiv:1812.09085v2, §4.
  • Use/classification: Context for the independently derived finite affine endpoint-cost arrangement. The library proves both signed factors, finite cuts and sign-cell cost-order invariance. The regression addresses only the fixed-push single-pencil claim; neither source's main matching-distance theorem is claimed formalized or refuted.
  • Rights/evidence: U; versioned primary HTML used for mathematical reading. No source prose, figures or code are reproduced.