https://arxiv.org/api/8aQMnxldVaJ0wV1+dehR0VYGdK02026-03-22T12:01:25Z58157515http://arxiv.org/abs/2602.06372v1Soft Bitopological Spaces via Soft Elements2026-02-06T04:00:38ZWe introduce soft bitopological spaces from the standpoint of soft elements. A soft bitopological space is a soft set equipped with two soft topologies. Following the classical construction of Goldar--Ray, each soft topology on $F$ induces an ordinary topology on the set $\SE(F)$ of soft elements; hence every soft bitopological space canonically determines a genuine bitopological space on $\SE(F)$. Within this setting we define pairwise soft separation axioms ($T_0$, $T_1$, $T_2$) and a notion of pairwise soft compactness, and we compare them with their parameterwise counterparts. For canonical (sectionwise generated) soft bitopologies, we show that the pairwise soft $T_i$ axioms are equivalent to the corresponding pairwise $T_i$ axioms on each parameter space. Compactness exhibits a finiteness phenomenon: when the parameter set is finite, componentwise pairwise compactness forces pairwise soft compactness, while an infinite-parameter example shows that the finiteness assumption is essential. Examples are included to clarify how the induced bitopology on $\SE(F)$ may behave differently from the original soft bitopology.2026-02-06T04:00:38ZS. Rayhttp://arxiv.org/abs/2602.04647v2On $κ$-Frechet-Urysohn topological groups2026-02-05T13:03:20ZWe characterize $κ$-Fréchet--Urysohn topological groups. Using this characterization we show that: (1) a hemicompact topological group is $κ$-Fréchet--Urysohn iff it is locally compact, and (2) if $F$ is a closed metrizable subspace of a topological vector space (tvs) $E$ such that the quotient $E/F$ is a $κ$-Fréchet--Urysohn space, then also $E$ is a $κ$-Fréchet--Urysohn space. Consequently, the product of a $κ$-Fréchet--Urysohn tvs and a metrizable tvs is a $κ$-Fréchet--Urysohn space. Under Martin's Axiom, we construct a countable Boolean $κ$-Fréchet--Urysohn group which is not a $k_{\mathbb R}$-space.2026-02-04T15:23:37ZSaak GabriyelyanAlexander V. OsipovEvgenii Reznichenkohttp://arxiv.org/abs/2602.03951v2Representation Geometry as a Diagnostic for Out-of-Distribution Robustness2026-02-05T04:54:34ZRobust generalization under distribution shift remains difficult to monitor and optimize in the absence of target-domain labels, as models with similar in-distribution accuracy can exhibit markedly different out-of-distribution (OOD) performance. While prior work has focused on training-time regularization and low-order representation statistics, little is known about whether the geometric structure of learned embeddings provides reliable post-hoc signals of robustness. We propose a geometry-based diagnostic framework that constructs class-conditional mutual k-nearest-neighbor graphs from in-distribution embeddings and extracts two complementary invariants: a global spectral complexity proxy based on the reduced log-determinant of the normalized Laplacian, and a local smoothness measure based on Ollivier--Ricci curvature. Across multiple architectures, training regimes, and corruption benchmarks, we find that lower spectral complexity and higher mean curvature consistently predict stronger OOD accuracy across checkpoints. Controlled perturbations and topological analyses further show that these signals reflect meaningful representation structure rather than superficial embedding statistics. Our results demonstrate that representation geometry enables interpretable, label-free robustness diagnosis and supports reliable unsupervised checkpoint selection under distribution shift.2026-02-03T19:13:36ZAli ZiaFarid Hazratianhttp://arxiv.org/abs/2510.07541v2Strongly bounded generation in transformation groups2026-02-05T00:45:10ZWord metrics on finitely generated groups have canonical quasi-isometry classes, making quasi-isometry invariants genuine group invariants. Rosendal generalized this phenomenon to topological groups through CB-generation, but in the general topological setting the resulting quasi-isometry invariants are not invariants of the underlying abstract group. Specializing to the discrete case yields what we call SB-generated groups, where the invariants are genuinely algebraic. We show that SB-generation arises naturally in transformation groups by identifying several broad families of examples: the identity component of homeomorphism groups of closed manifolds, certain big mapping class groups, and homeomorphism groups of compact well-ordered spaces with successor limit capacity. These results demonstrate that SB-generation provides a robust extension of finite generation.2025-10-08T20:56:42Z22 pages, 1 figure, v2: fixed typesetting issue with cleverref, correcting referencing of propositions, lemmas, and corollaries: no mathematical changeNicholas G. Vlamishttp://arxiv.org/abs/2408.07316v2Relative sectional number and the coincidence property2026-02-04T22:48:37ZFor a Hausdorff space $Y$, a topological space $X$ and a map $g:X\to Y$, we present a connection between the relative sectional number of the first coordinate projection $π_{2,1}^Y:F(Y,2)\to Y$ with respect to $g$, and the coincidence property (CP) for $(X,Y;g)$, where $F(Y,2)$ stands for the ordered configuration space of $2$ distinct points on $Y$, and $(X,Y;g)$ has the coincidence property (CP) if, for every map $f:X\to Y$, there is a point $x$ of $X$ such that $f(x)=g(x)$. Explicitly, we demonstrate that $(X,Y;g)$ has the CP if and only if 2 is the minimal cardinality of open covers $\{U_i\}_{1\leq i\leq n}$ of $X$ such that each $U_i$ admits a local lifting for $g$ with respect to $π_{2,1}^Y$. This characterization connects a standard problem in coincidence theory to current research trends in sectional category and topological robotics. Motivated by this connection, we introduce the notion of relative topological complexity of a map.2024-08-14T06:25:36Z16 pages. Final versionTopological Methods in Nonlinear Analysis. Online. 11 December 2025. Vol. 66, no. 2, pp. 479 - 496Cesar A. Ipanaque ZapataFelipe A. Torres Estrella10.12775/TMNA.2025.016http://arxiv.org/abs/2602.23381v1Universality of Shallow and Deep Neural Networks on Non-Euclidean Spaces2026-02-03T17:46:46ZWe develop a framework for shallow and deep neural networks whose inputs range over a general topological space. The model is built from a prescribed family of continuous feature maps and a fixed scalar activation function, and it reduces to multilayer feedforward networks in the Euclidean case. We focus on the universal approximation property and establish general conditions under which such networks are dense in spaces of continuous vector-valued functions on arbitrary and locally convex topological spaces. In the absence of width constraints, we obtain universality results that extend classical approximation theorems to non-Euclidean settings. A central focus of the paper is the deep narrow framework, in which the width of each hidden layer is uniformly bounded while the depth is allowed to grow. We identify conditions under which such width constrained deep networks retain universal approximation power. As a concrete example, we employ Ostrand's extension of the Kolmogorov superposition theorem to derive an explicit universality result for products of compact metric spaces, with width bounds expressed in terms of topological dimension.2026-02-03T17:46:46Z23 pages, 35 referencesVugar Ismailovhttp://arxiv.org/abs/2502.17707v4On non-Hausdorff manifolds2026-02-01T21:00:37ZIn this long note, we investigate various purely topological aspects of non-Hausdorff manifolds (NH-manifolds for short). Our emphasis is on manifolds which exhibit homogeneity or weakenings thereof, in particular being everywhere non-Hausdorff. Homogeneous NH-manifolds and everywhere non-Hausdorff manifolds are respectively called HNH- and ENH-manifolds. We write $NH_X(x)$ for the subset of points of a space $X$ which cannot be separated of $x$ by open sets. The topics covered in this note are the following.
-- General (basic) properties of manifolds and their quasi-compact or quasi-countably compact subspaces.
-- Covering properties implying the Hausdorffness of (weakly) homogeneous manifolds.
-- (Non-)existence of hereditarily separable ENH-manifolds (sometimes under set theoretic hypotheses).
-- Non-existence of a quasi-countably compact ENH-manifold.
-- Properties of NH-manifolds which imply that $NH_M(x)$ is discrete, or at least ``simple''.
-- Constructions of HNH-manifolds such that $NH(x)$ is non-homogeneous, for instance a countable union of closed intervals and $n$-torii.
-- Constructions of NH-manifolds $M$ with a point $x$ such that $NH_M(x)$ is homeomorphic to various ``complicated'' spaces, in particular in dimension $1$ and $2$.
We use elementary (or at least well known) methods of general or set theoretic topology, with a little bit of conformal theory and dynamical systems (flows) in some constructions. Many pictures are given to illustrate the constructions, and the proofs are rather detailed, which is the main reason for the length of this note.2025-02-24T23:00:04ZWork in progress whose contents might change over time. Some pictures use colors and can be difficult to decipher for color blind readers. V4: See the list of main changes on first page. Also, typos corrected and improved presentation in many placesMathieu Baillifhttp://arxiv.org/abs/2510.25200v2On Quasi-Modular Pseudometric Spaces and Asymmetric Uniformities2026-01-31T22:23:51ZWe study quasi-modular pseudometric spaces as asymmetric refinements of modular metric structures. To each such space we associate canonical forward and backward quasi-uniformities and the corresponding directional topologies.
We introduce directional notions of convergence, completeness, total boundedness, and compactness, and show that these properties are not preserved under symmetrization. In particular, forward and backward completeness may differ, and compactness of the symmetrized uniformity does not imply directional compactness.
Using enriched category theory as a comparison framework, we show that symmetrization yields a symmetric enriched category whose Cauchy completion coincides with the classical uniform completion, while directional notions remain invisible at this level.2025-10-29T06:10:52Z13 pagesPhilani Rodney Majozihttp://arxiv.org/abs/2602.00645v1Best Proximity Point Results for Perimetric Contractions2026-01-31T10:33:46ZThis paper has two aims, first one is to introduce special kind of proximal contractions guaranteeing a finite number of best proximity points, and second one is to derive best proximity point results for perimetric contractions. To meet these two aims, we introduce two new proximal contractions: perimetric proximal contractions of the first and the second kind, and derive best proximity point results for these mappings. We establish that for these particular mappings, best proximity points are not necessarily unique; however, we provide an upper bound, proving that at most two such points can exist. To establish the validity of our results, we provide illustrative examples demonstrating that these newly defined mappings can possess unique or exactly two best proximity points.2026-01-31T10:33:46Z10 pagesHiranmoy GaraiEvgeniy PetrovPratikshan MondalLakshmi Kanta Deyhttp://arxiv.org/abs/2506.15422v4Super-Dense Sets and Their Role in the Theory of Normal Numbers2026-01-30T20:57:33ZWe introduce and study a new topological notion of the size for subsets of the real line, called \emph{super-density}. A set $A\subset\mathbb{R}$ is super-dense if for every non-empty open interval $I$ and every nowhere constant continuous function $\varphi\colon I\to\mathbb{R}$, we have $\varphi(I\cap A)\cap A\neq\emptyset$. We first establish basic properties of super-dense sets. Our main topological result characterizes them within the framework of Baire category: a set with the Baire property is super-dense if and only if it is co-meager. We then investigate the implications for the theory of normal numbers. We prove that the set of non-normal numbers is super-dense, whereas the set of normal numbers is not. Consequently, no nowhere constant continuous function can map all non-normal numbers to normal numbers. Conversely, we explicitly construct a computable nowhere constant continuous function that maps all normal numbers to non-normal numbers. Finally, we provide a constructive algorithm that, given any countable family of nowhere constant continuous functions, produces a real number $x$ such that $x$ and all its images under these functions are non-normal. As a corollary, we obtain the existence of a non-normal number $x$ such that $e^{αx}$ is non-normal for every non-zero algebraic $α$.2025-06-18T12:48:17Z25 pages, restructured presentation with larger emphasis on topological results which have been expandedChokri Manaihttp://arxiv.org/abs/2601.01681v2Tameness of actions on finite rank median algebras2026-01-30T18:24:36ZWe prove that for (compact) finite-rank median algebras the geometric rank equals the independence number of all (continuous) median-preserving functions to $[0,1]$. Combined with Rosenthal's dichotomy, this yields a generalized Helly selection principle: for finite-rank median algebras, the space of all median-preserving functions to $[0,1]$ is sequentially compact in the pointwise topology. Generalizing joint results with E. Glasner on dendrons (rank-1), we establish that every continuous action of a topological group $G$ by median automorphisms on a finite-rank compact median algebra is Rosenthal representable, hence dynamically tame. As an application, the Roller-Fioravanti compactification of finite-rank topological median $G$-algebras with compact intervals is often a dynamically tame $G$-system.2026-01-04T22:30:31Z18 pagesMichael Megrelishvilihttp://arxiv.org/abs/2601.02279v2Hyperconvexity in partial metric spaces: challenges and outlooks2026-01-30T17:23:32ZIn this article, we present several different ways to define hyperconvexity in partial metric spaces. In particular, we show that the analogue of the Aronszajn--Panitchpakdi notion of hyperconvexity fails to exhibit certain key properties present in the classical metric setting.2026-01-05T17:10:17Z10 pagesDariusz BugajewskiPiotr KasprzakOlivier Olela-Otafuduhttp://arxiv.org/abs/2602.23380v1Reeb spaces of functions being analytic on dense subsets and their graph structures2026-01-30T01:15:16ZReeb spaces of real-valued functions on manifolds are the spaces of all connected components (contours) of level sets and endowed with the natural quotient topology. They have been fundamental and strong tools in investigating manifolds via smooth functions with mild critical points since the birth of fundamental theory of Morse functions in the 20th century.
We are concerned with topologies and combinatorics of them. Following an explicit note on explicit Reeb spaces of explicit functions which are real analytic (on dense sets) and seem to be simplest and most fundamental, edited by the author himself.
We investigate other construction of examples of such functions and their Reeb spaces. Reeb spaces are naturally graphs in considerable cases and as another work, we also discuss natural definitions of vertices for them.2026-01-30T01:15:16Z9 pagesNaoki Kitazawahttp://arxiv.org/abs/2511.01273v2Homeomorphisms of continua through projective Fraïssé limits2026-01-29T00:32:57ZWe study homeomorphisms and the homeomorphism groups of compact metric spaces using the automorphism groups of projective Fraïssé limits. In our applications, we investigate the Polish group ${\rm Homeo}(P)$ of all homeomorphisms of the pseudoarc $P$ using the automorphism group ${\rm Aut}(\mathbb{P})$ of the pre-pseudoarc $\mathbb{P}$. Strengthening results from the literature, we show that the diagonal conjugacy action of ${\rm Homeo}(P)$ on ${\rm Homeo}(P)^{\mathbb{N}}$ has a dense orbit. In our second application, we show that there exists a homeomorphism of $P$ that is not conjugate in ${\rm Homeo}(P)$ to an element of ${\rm Aut}(\mathbb{P})$.2025-11-03T06:47:09ZMárk PoórSławomir Soleckihttp://arxiv.org/abs/2508.14651v2A number of properties enjoyed by two specially constructed topologies on $C(X)$2026-01-28T08:34:20ZIf $I$ is an ideal in the ring $C(X)$ of all real valued continuous functions defined over a Tychonoff space $X$, then $X$ is called $I$-$pseudocompact$ if the set $X\setminus \bigcap Z[I]$ is a bounded subset of $X$. Corresponding to $I$, the $m^I$-topology and $u^I$-topology on $C(X)$, generalizing the well-known $m$-topology and $u$-topology in $C(X)$ respectively are already there in the literature. It is proved amongst others that the $m^I$-topology is first countable if and only if the $u^I$-topology= $m^I$-topology on $C(X)$ if and only if $X$ is $I$-$pseudocompact$. A special case of this result on choosing $I=C(X)$ reads: the $u$-topology and $m$-topology on $C(X)$ coincide if and only if $X$ is pseudocompact. It is established that the $m^I$-topology on $C(X)$ is second countable if and only if it is $\aleph_0$-$bounded$ if and only if $X$ is compact, metrizable and $I=C(X)$. Furthermore it is realized that the $m^I$ topology on $C(X)$ is hemicompact if and only if it is $σ$-compact if and only if this topology is $H$-$bounded$ if and only if $X$ is finite and $I=C(X)$.2025-08-20T12:08:29ZSoumajit DeySudip Kumar AcharyyaDhananjoy Mandal