https://arxiv.org/api/A5dbFSI1kfiNM2psapWMRhQWxFk 2026-03-20T08:58:15Z 17660 30 15 http://arxiv.org/abs/2603.16123v1 2026-03-17T05:08:30Z

Neural networks systematically fail at compositional generalization -- producing correct outputs for novel combinations of known parts. We show that this failure is architectural: compositional generalization is equivalent to functoriality of the decoder, and this perspective yields both guarantees and impossibility results. We compile Higher Inductive Type (HIT) specifications into neural architectures via a monoidal functor from the path groupoid of a target space to a category of parametric maps: path constructors become generator networks, composition becomes structural concatenation, and 2-cells witnessing group relations become learned natural transformations. We prove that decoders assembled by structural concatenation of independently generated segments are strict monoidal functors (compositional by construction), while softmax self-attention is not functorial for any non-trivial compositional task. Both results are formalized in Cubical Agda. Experiments on three spaces validate the full hierarchy: on the torus ($\mathbb{Z}^2$), functorial decoders outperform non-functorial ones by 2-2.7x; on $S^1 \vee S^1$ ($F_2$), the type-A/B gap widens to 5.5-10x; on the Klein bottle ($\mathbb{Z} \rtimes \mathbb{Z}$), a learned 2-cell closes a 46% error gap on words exercising the group relation.

2026-03-17T05:08:30Z 20 pages, 10 tables. Code and Cubical Agda formalization: https://github.com/karsar/hott_neuro Karen Sargsyan http://arxiv.org/abs/2603.15873v1 2026-03-16T20:05:07Z

We construct multiplicative norms on equivariant nonconnective algebraic $K$-theory for finite groups $G$. We also construct a genuine equivariant version of THH equipped with a Dennis trace map from K-theory compatible with the multiplicative norms. To do so, we follow the general strategy of Blumberg-Gepner-Tabuada in the nonequivariant case by generalizing their category of localizing motives to the genuine equivariant context, building upon the theory of perfect $G$-stable categories of the first-named author. Crucially, we proceed using the recent perspective on noncommutative motives by the second-named author with Sosnilo and Winges which allows us to deal with non-exact functors on this category of motives. Together with an isotropy separation argument for equivariant cubes, we prove our main theorem that norms of stable categories preserve equivariant motivic equivalences. As an immediate consequence, we obtain a unique equivariant multiplicative refinement of nonconnective algebraic $K$-theory. From these constructions and results, we draw several applications, namely: (1) that the endofunctor of (equivariant) tensor powers on ordinary perfect stable categories preserve motivic equivalences; (2) that the multiplicative norms also preserve the additive motivic equivalences, thus yielding a motivic refinement of a result of Elmanto-Haugseng and Cnossen-Haugseng-Lenz-Linskens that connective algebraic K-theory admits multiplicative norms; (3) we construct a genuine equivariant version of topological Hochschild homology equipped with a Dennis trace map that is compatible with multiplicative norms; and (4) we prove that every genuine $G$-spectrum is the K-theory of a perfect $G$-stable category.

2026-03-16T20:05:07Z 50 pages. Comments very welcome! Kaif Hilman Maxime Ramzi http://arxiv.org/abs/2202.02591v3 2026-03-16T20:05:00Z

For an atomic orbital base category in the sense of Barwick-Dotto-Glasman-Nardin-Shah, we introduce the category of parametrised perfect-stable categories and use it to construct the parametrised version of noncommutative motives in which algebraic K-theory is corepresented. Furthermore, we initiate a rudimentary theory of parametrised cubes which could be of independent interest, generalising some of the elements in Dotto's theory of equivariant Goodwillie calculus beyond the equivariant case. Using this cubical theory, we show that in the equivariant case for finite 2-groups G, the parametrised noncommutative motives canonically refine to G-symmetric monoidal categories. Consequently, this endows the equivariant algebraic K-theory spectra for these groups with the structure of E-infinity-ring spectra equipped with multiplicative norms in the sense of Hill-Hopkins-Ravenel. Along the way, we will also provide a machine to manufacture G-symmetric monoidal categories from symmetric monoidal categories equipped with G-actions and elucidate how the aforementioned parametrised perfect-stable categories relate to Mackey functors valued in perfect-stable categories.

2022-02-05T16:17:14Z 58 pages; v3 minor corrections and changes in exposition, addition of Hypotheses 2.2.17 and 4.2.1 Kaif Hilman http://arxiv.org/abs/2603.15384v1 2026-03-16T14:58:32Z

We improve and extend persistence spheres, introduced in~\cite{pegoraro2025persistence}. Persistence spheres map an integrable measure $μ$ on the upper half-plane, including persistence diagrams (PDs) as counting measures, to a function $S(μ)\in C(\mathbb{S}^2)$, and the map is stable with respect to 1-Wasserstein partial transport distance $\mathrm{POT}_1$. Moreover, to the best of our knowledge, persistence spheres are the first explicit representation used in topological machine learning for which continuity of the inverse on the image is established at every compactly supported target. Recent bounded-cardinality bi-Lipschitz embedding results in partial transport spaces, despite being powerful, are not given by the kind of explicit summary map considered here. Our construction is rooted in convex geometry: for positive measures, the defining ReLU integral is the support function of the lift zonoid. Building on~\cite{pegoraro2025persistence}, we refine the definition to better match the $\mathrm{POT}_1$ deletion mechanism, encoding partial transport via a signed diagonal augmentation. In particular, for integrable $μ$, the uniform norm between $S(0)$ and $S(μ)$ depends only on the persistence of $μ$, without any need of ad-hoc re-weightings, reflecting optimal transport to the diagonal at persistence cost. This yields a parameter-free representation at the level of measures (up to numerical discretization), while accommodating future extensions where $μ$ is a smoothed measure derived from PDs (e.g., persistence intensity functions~\citep{wu2024estimation}). Across clustering, regression, and classification tasks involving functional data, time series, graphs, meshes, and point clouds, the updated persistence spheres are competitive and often improve upon persistence images, persistence landscapes, persistence splines, and sliced Wasserstein kernel baselines.

2026-03-16T14:58:32Z Matteo Pegoraro http://arxiv.org/abs/2603.08653v2 2026-03-16T14:23:50Z

In this paper we prove the theorem of the heart for Weibel's homotopy $K$-theory $KH.$ Namely, if $\mathcal{C}$ is a small stable $\infty$-category with a bounded $t$-structure, then the realization functor $D^b(\mathcal{C}^{\heartsuit})\to \mathcal{C}$ induces an equivalence of spectra $KH(\mathcal{C}^{\heartsuit})\xrightarrow{\sim}KH(\mathcal{C}).$ In a certain sense this result is dual to the Dundas-Goodwillie-McCarthy theorem. We deduce the dévissage theorem for $KH$ of abelian categories, also on the level of spectra (in all degrees). More generally, we prove these results for dualizable categories with nice $t$-structures and for the so-called coherently assembled abelian categories. The proof is heavily based on another new result, which is a much stronger version of Barwick's theorem of the heart. Its special case states the following: if $\mathcal{C}$ is a small stable category with a bounded $t$-structure, such that for some $n\geq 1$ the realization functor induces isomorphisms on $\operatorname{Ext}^{\leq n}$ between the objects of $\mathcal{C}^{\heartsuit},$ then the map $K_j(\mathcal{C}^{\heartsuit})\to K_j(\mathcal{C})$ is an isomorphism for $j\geq -n-1,$ and a monomorphism for $j = -n-2.$ Moreover, we prove that these estimates are sharp, even for dg categories over a field. In particular the naive $K$-theoretic theorem of the heart fails for $K_{-3}.$

2026-03-09T17:31:36Z 93 pages; v2: minor changes, typos fixed, references added Alexander I. Efimov http://arxiv.org/abs/2603.15303v1 2026-03-16T14:03:19Z

Alesker's theory of generalized valuations unifies smooth measures and constructible functions on real analytic manifolds, extending classical operations on functions and measures. Alesker showed that these operations agree with the sheaf-theoretic ones on constructible functions under restrictive assumptions, leaving key aspects conjectural. In this article, we close this gap by proving that the two approaches indeed coincide on constructible functions under mild transversality assumptions. Our proof is based on a comparison with the corresponding operations on characteristic cycles. As applications, we extend additive kinematic formulas from convex bodies to compact subanalytic sets in Euclidean spaces and derive new kinematic formulas on the 3-sphere.

2026-03-16T14:03:19Z A first version of this work appeared on HAL here: https://hal.science/hal-05011973. This second version contains improved assumptions on main theorems and applications to kinematic formulas Andreas Bernig Vadim Lebovici http://arxiv.org/abs/2603.15165v1 2026-03-16T11:59:37Z

The problem of graph burning was firstly introduced as a model for different processes of social and network interactions. Recently, the authors of the present paper developed methods of algebraic topology for investigation of this problem. This approach is based on the new definition of burning process which excludes the possibility to choose at any moment vertex for burning from the set of vertices which are already burned at this moment. In this paper we continue to study such burning process using algebraic topology methods. We prove the result about relations between burnings of a graph and burnings of its spanning trees that is similar to the classical case. Afterwards, we describe properties of trees burnings. In particular, we prove that a burning of a tree defines a structure of a digraph on the tree and investigate this structure. We introduce and study a strong burning configuration space of a graph and new strong burning homology which are similar to burning homology defined in our previous paper, but arise from burning homomorphism.

2026-03-16T11:59:37Z 23 pages, 6 figures, 1 table Yuri Muranov Anna Muranova http://arxiv.org/abs/2506.23348v2 2026-03-16T11:57:59Z

We prove that the space of shifted Poisson structures on a derived scheme $X$ locally of finite presentation is equivalent to the space of shifted Lagrangian thickenings out $X$, solving a conjecture in shifted Poisson geometry. As a corollary, we show that for $M$ a compact oriented $d$-dimensional manifold and an $n$-shifted Poisson structure on $X$, the mapping stack $\mathrm{Map}(M,X)$ has an $(n-d)$-shifted Poisson structure. It extends a known theorem for shifted symplectic structures to shifted Poisson structures.

2025-06-29T17:50:06Z 50 pages. Revised version. Any comments are welcome! Nikola Tomić http://arxiv.org/abs/2603.15079v1 2026-03-16T10:33:34Z

Persistent homology (PH) -- the conventional method in topological data analysis -- is computationally expensive, requires further vectorization of its signatures before machine learning (ML) can be applied, and captures information along only the spatial axis. For time series data, we propose Euler Characteristic Surfaces (ECS) as an alternative topological signature based on the Euler characteristic ($χ$) -- a fundamental topological invariant. The ECS provides a computationally efficient, spatiotemporal, and inherently discretized feature representation that can serve as direct input to ML models. We prove a stability theorem guaranteeing that the ECS remains stable under small perturbations of the input time series. We first demonstrate that ECS effectively captures the nontrivial topological differences between the limit cycle and the strange attractor in the Rössler system. We then develop an ECS-based classification framework and apply it to five benchmark biomedical datasets (four ECG, one EEG) from the UCR/UEA archive. On $\textit{ECG5000}$, our single-feature ECS classifier achieves $98\%$ accuracy with $O(n+R\cdot T)$ complexity, compared to $62\%$ reported by a recent PH-based method. An AdaBoost extension raises accuracy to $98.6\%$, matching the best deep learning results while retaining full interpretability. Strong results are also obtained on $\textit{TwoLeadECG}$ ($94.1\%$) and $\textit{Epilepsy2}$ ($92.6\%$).

2026-03-16T10:33:34Z Salam Rabindrajit Luwang Sushovan Majhi Vishal Mandal Atish J. Mitra Md. Nurujjaman Buddha Nath Sharma http://arxiv.org/abs/2603.15010v1 2026-03-16T09:15:52Z

Thom polynomials are universal cohomological obstructions to the appearance of singularities of given types in differentiable maps. As an application, various invariants of immersions have been expressed in terms of singularities of extensions of immersions (known as singular Seifert surfaces). To place these results in a unified framework, we aim in this paper to establish the foundation of a relative version of Thom polynomial theory. Our result consists of three parts. (1) We introduce the notion of relative Thom polynomials, which are relative cohomological obstructions for extensions of prescribed maps between codimension-zero submanifolds that avoid singularities of given types. (2) We show a structure theorem for relative Thom polynomials when the prescribed map is a framed immersion. It expresses them as the sum of the naive substitution of Kervaire's relative characteristic classes into the absolute Thom polynomial and a correction term. As a consequence, the correction term forms a regular homotopy invariant of the prescribed map. (3) We determine correction terms in several cases, not only reinterpreting earlier works as instances of relative Thom polynomials but also applying our framework to the type $A_1$. We observe that these terms vanish or consist of classical invariants and their variants.

2026-03-16T09:15:52Z 32 pages, 2 tables Masato Tanabe http://arxiv.org/abs/2603.14788v1 2026-03-16T03:38:37Z

For positive integers $k$, $n$, and $g$ with $k\geq2$, we give a closed-form expression for the $k$-th $\mathbb{Z}_2$-zero-divisor cup length $\mathsf{zcl}_k(SP^n(N_g))$ of the $n$-th symmetric product $SP^n(N_g)$ of the closed non-orientable surface $N_g$ of genus $g$. This allows us to estimate, and in some cases, completely determine, the $k$-th sequential topological complexity $\mathsf{TC}_k(SP^n(N_g))$, as well as the Lusternik--Schnirelmann category of the homotopy cofiber of the $k$-th diagonal map $SP^n(N_g) \to (SP^n(N_g))^k$. Our results recover previously known facts for even-dimensional real projective spaces ($g=1$) and closed non-orientable surfaces ($n=1$). In addition, we show that, as $g$ grows, $\mathsf{TC}_2(SP^n(N_g))$ behaves in a different way as all other invariants $\mathsf{TC}_k(SP^n(N_g))$ do. Likewise, as $k$ grows, we describe an eventual maximal-possible linear growth of $\mathsf{zcl}_k(SP^n(N_g))$, which allows us to prove the rationality conjecture of Farber and Oprea for the TC-generating function of $SP^n(N_g)$.

2026-03-16T03:38:37Z 29 pages. Comments welcome! Jesús González Ekansh Jauhari http://arxiv.org/abs/2303.10822v6 2026-03-16T03:28:43Z

This paper considers non-Abelian homology groups of a group diagram introduced as homotopy groups of a simplicial change. We prove a theorem stating that the non-Abelian homology groups of a group diagram are isomorphic to the homotopy groups of the homotopy colimit of a classifying space diagram, with the dimension shifted by 1. Bousfield and Kan proved an isomorphism between the homotopy groups of an Abelian simplicial group and the homology groups of this simplicial group. We generalize this to non-Abelian simplicial groups. We also develop a method for finding a non-zero homotopy group of smallest dimension for the homotopy colimit of classifying spaces. For a group diagram over a free category with a zero colimit, we obtain a criterion for the isomorphism of the first non-Abelian and Abelian homology groups.

2023-03-20T01:37:01Z 26 pages Ahmet A. Husainov http://arxiv.org/abs/2603.14440v1 2026-03-15T15:32:26Z

The electric Gauss law in 11D SuGra is famously non-linear, whence its flux quantization must be in nonabelian cohomology. We have previously shown that the minimal admissible choice is 4-Cohomotopy, which in the presence of magnetized M5-probes takes its relative twistorial form. Here we discuss how this situation is further refined in the presence of M-string probes on the M5-worldvolume. Based on the superspace formulation of 11D SuGra, we find the nested Bianchi identities by iterating the superembedding construction for super p-branes. The resulting probe brane hierarchy (M1 on magnetized M5 in 11D bulk) turns out to admit flux quantization in a doubly-relative form of twisted Cohomotopy, classified by the factorization of the quaternionic Hopf fibration through the twistor fibration. The further equivariant refinement of this cohomology theory reduces on A-type singularities to a form of relative 2-Cohomotopy which geometrically engineers Chern-insulator phases on $\mathrm{M5}\cap \mathrm{A}_n$, with the M-string playing the role of gapped nodal lines.

2026-03-15T15:32:26Z 25 pages, 4 figures Pinak Banerjee Hisham Sati Urs Schreiber http://arxiv.org/abs/2511.09189v4 2026-03-15T04:59:28Z

Any $C^*$-algebra can be regarded as a generalization of locally compact, Hausdorff topological space $\mathcal X$. From the commutative commutative Gelfand-Naĭmark theorem it follows that the spectrum of any commutative $C^*$-algebra is a locally compact, Hausdorff space which have the exact information of the $C^*$-algebra. Here we consider a Gelfand spaces of $C^*$-algebras which can be regarded as a generalization of the spectrum. In case of commutative $C^*$-algebras the Gelfand space coincides with the spectrum. Generally Gelfand spaces are not Hausdorff and provide more detailed information of noncommutative $C^*$-algebras. Sometimes the Gelfand space contains the full information of noncommutative $C^*$-algebra. Usage of Gelfand spaces of $C^*$-algebra enables us to define some $C^*$-algebraic analogs of several notions of the classical algebraic topology.

2025-11-12T10:37:33Z 111 pages, 59 references Petr Ivankov http://arxiv.org/abs/2603.18041v1 2026-03-14T21:47:16Z

Swarm and constellation reconfiguration can be viewed as motion of an unordered point configuration in an ambient space. Here, we provide persistence-stable, symmetry-invariant geometric representations for comparing and monitoring multi-agent configuration data. We introduce a quotient formation space $\mathcal{S}_n(M,G)=M^n/(G\times S_n)$ and a formation matching metric $d_{M,G}$ obtained by optimizing a worst-case assignment error over ambient symmetries $g\in G$ and relabelings $σ\in S_n$. This metric is a structured, physically interpretable relaxation of Gromov--Hausdorff distance: the induced inter-agent metric spaces satisfy $d_{\mathrm{GH}}(X_x,X_y)\le d_{M,G}([x],[y])$. Composing this bound with stability of Vietoris--Rips persistence yields $d_B(Φ_k([x]),Φ_k([y]))\le d_{M,G}([x],[y])$, providing persistence-stable signatures for reconfiguration monitoring. We analyze the metric geometry of $(\mathcal{S}_n(M,G),d_{M,G})$: under compactness/completeness assumptions on $M$ and compact $G$ it is compact/complete and the metric induces the quotient topology; if $M$ is geodesic then the quotient is geodesic and exhibits stratified singularities along collision and symmetry strata, relating it to classical configuration spaces. We study expressivity of the signatures, identifying symmetry-mismatch and persistence-compression mechanisms for non-injectivity. Finally, in a phase-circle model we prove a conditional inverse theorem: under semicircle support and a gap-labeling margin, the $H_0$ signature is locally bi-Lipschitz to $d_{M,G}$ up to an explicit factor, yielding two-sided control. Examples on $\mathbb{S}^2$ and $\mathbb{T}^m$ illustrate satellite-constellation and formation settings.

2026-03-14T21:47:16Z 20 pages Mark M. Bailey