Relative Obstructions and Spectral Diagnostics for Sheaves on Cell Complexes
Fuente:
arXiv
Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866908812620857344 |
|---|---|
| author | Yokoyama, Shinobu |
| author_facet | Yokoyama, Shinobu |
| contents | Many structured systems admit locally consistent descriptions that nevertheless fail to globalize when constrained by an ambient reference or feasibility condition. Diagnosing such failures is naturally an evaluative problem: given a fixed model and a grounding, can one determine whether they are structurally compatible, and if not, identify the nature and localization of the obstruction?
In this work, we introduce a sheaf-theoretic and spectral framework for evaluating structural inconsistency as a \emph{relative} phenomenon. A model is represented by a cellular sheaf $\mathcal F$ on a cell complex, together with a morphism into a grounding sheaf $\mathcal W$ encoding admissible global behavior. Failure of compatibility is captured by the mapping cone of this morphism, whose cohomology computes the relative groups $H^*(K;\mathcal F,\mathcal W)$ and separates intrinsic obstructions from inconsistencies induced by the grounding.
Beyond exact cohomological classification, we develop \emph{spectral witnesses} derived from regular and mapping-cone Laplacians. The spectra of these operators provide computable, quantitative indicators of inconsistency, encoding both robustness and spatial localization through spectral gaps, integrated energies, and eigenmode support. These witnesses enable comparison of distinct inconsistency mechanisms in fixed systems without learning, optimization, or modification of the underlying representation.
The proposed framework is domain-agnostic and applies to a broad class of structured models where feasibility is enforced locally but evaluated globally. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_19056 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Relative Obstructions and Spectral Diagnostics for Sheaves on Cell Complexes Yokoyama, Shinobu Algebraic Topology 55N30 Many structured systems admit locally consistent descriptions that nevertheless fail to globalize when constrained by an ambient reference or feasibility condition. Diagnosing such failures is naturally an evaluative problem: given a fixed model and a grounding, can one determine whether they are structurally compatible, and if not, identify the nature and localization of the obstruction? In this work, we introduce a sheaf-theoretic and spectral framework for evaluating structural inconsistency as a \emph{relative} phenomenon. A model is represented by a cellular sheaf $\mathcal F$ on a cell complex, together with a morphism into a grounding sheaf $\mathcal W$ encoding admissible global behavior. Failure of compatibility is captured by the mapping cone of this morphism, whose cohomology computes the relative groups $H^*(K;\mathcal F,\mathcal W)$ and separates intrinsic obstructions from inconsistencies induced by the grounding. Beyond exact cohomological classification, we develop \emph{spectral witnesses} derived from regular and mapping-cone Laplacians. The spectra of these operators provide computable, quantitative indicators of inconsistency, encoding both robustness and spatial localization through spectral gaps, integrated energies, and eigenmode support. These witnesses enable comparison of distinct inconsistency mechanisms in fixed systems without learning, optimization, or modification of the underlying representation. The proposed framework is domain-agnostic and applies to a broad class of structured models where feasibility is enforced locally but evaluated globally. |
| title | Relative Obstructions and Spectral Diagnostics for Sheaves on Cell Complexes |
| topic | Algebraic Topology 55N30 |
| url | https://arxiv.org/abs/2601.19056 |