Margin in Abstract Spaces

2026-03-07T14:02:39Z

Margin-based learning, exemplified by linear and kernel methods, is one of the few classical settings where generalization guarantees are independent of the number of parameters. This makes it a central case study in modern highly over-parameterized learning. We ask what minimal mathematical structure underlies this phenomenon. We begin with a simple margin-based problem in arbitrary metric spaces: concepts are defined by a center point and classify points according to whether their distance lies below $r$ or above $R$. We show that whenever $R>3r$, this class is learnable in \emph{any} metric space. Thus, sufficiently large margins make learnability depend only on the triangle inequality, without any linear or analytic structure. Our first main result extends this phenomenon to concepts defined by bounded linear combinations of distance functions, and reveals a sharp threshold: there exists a universal constant $γ>0$ such that above this margin the class is learnable in every metric space, while below it there exist metric spaces where it is not learnable at all. We then ask whether margin-based learnability can always be explained via an embedding into a linear space -- that is, reduced to linear classification in some Banach space through a kernel-type construction. We answer this negatively by developing a structural taxonomy of Banach spaces: if a Banach space is learnable for some margin parameter $γ\geq 0$, then it is learnable for all such $γ$, and in infinite-dimensional spaces the sample complexity must scale polynomially in $1/γ$. Specifically, it must grow as $(1/γ)^p$ for some $p\ge 2$, and every such rate is attainable.

The speed measure and absolute continuity for curves in metric spaces

2026-03-07T09:27:00Z

We define the speed measure $ν$ for mappings $γ:I\to X$ from an interval to a metric space that are locally of bounded variation. We characterize continuity and absolute continuity of $γ$ in terms of $ν$ and identify the Radon-Nikodým derivative of $ν$ with respect to Lebesgue measure as the metric speed of $γ$. In doing so we prove an extension of the Banach-Zaretsky theorem.

On the weakest conditions for the existence of fixed points of Kannan and Chatterjea type contractions

2026-03-07T01:07:37Z

In this paper, we study the weakest possible conditions for fixed point theorems involving two classes of mappings defined by Kannan and Chatterjea. Our approach relies on the so-called CJM condition, which was originally introduced by Ćirić [5], and later, Suzuki [18] showed that the CJM condition is necessary for the existence of fixed points and the convergence of all Picard sequences of Banach type mappings. Our aim is to extend Suzuki's approach to the case of Kannan and Chatterjea mappings. In particular, in the first case, we discuss the equivalence of previously known conditions and establish that our conditions are optimal for ensuring that all Picard sequences converge to a fixed point of a mapping.

Raja's covering index of $L_p$ spaces

2026-03-06T22:41:01Z

We study Raja's covering index $Θ_X(n)$ for classical $L_p$-spaces and their non-commutative counterparts. For infinite-dimensional Hilbert spaces we compute the covering index exactly, proving \[ Θ_H(n)=n^{-1/2}\qquad(n\in\mathbb N); \] in particular $Θ_H(2)=1/\sqrt2$, thus answering a question of Raja about the precise two-piece covering index of $\elltwo$. For scalar-valued Lebesgue spaces $L_p(μ)$, $1\le p<\infty$, we construct an explicit block decomposition of the unit ball yielding the upper bound $Θ_{L_p(μ)}(n)\le n^{-1/p}$ for all $n\in\mathbb{N}$; in particular $Θ_{\ell_p}(n)\le n^{-1/p}$. For $1

Abstract fractional linear transformations

2026-03-06T22:25:11Z

We begin with (densely-defined) fractional linear transformations (FLT) on (some) Banach algebras and their relatives. This leads to Wedderburn's continued fractions (recursively-defined noncommutative polynomials) for any ring. Along the way, we discover a one-parameter family of (noncommutative) polynomials \st if one of them is invertible, then read in the opposite order, the corresponding polynomial is also invertible (extending the well known $1+ab$ is invertible if $1+ ba$ is, and the not-so-well-known, $a + abc + c$ and $a + cba + c$). This in turn leads to a definition of FLT for general rings $R$, which turns out to be PE$(2,R)$ (the projective elementary group). Using Wedderburn's polynomials, this permits us to define a length function on PE$(2,R)$, which suggests a stable range type condition (for $n =1$, it {\it is\/} stable range one, but higher values do not correspond. Again using the length results, we prove the expected results for PE$(2,R)$: under very modest conditions on $R$, the commutator subgroup of PE$(2,R)$ is perfect and of index one or two. Along the same lines, we also prove results on simplicity of the commutator subgroup: we require the usual generative properties on the simple ring $R$, as well either the very restrictive $1$ in the range, or a mild condition about invertibles, involving intersections of three translates of GL$(1,R)$. This last property is explored in the appendices, which give examples (and non-examples). Numerous questions suggest themselves throughout.

The Stockwell transform on Gelfand pairs and localization operators

2026-03-06T22:17:14Z

This paper addresses the extension of the Stockwell transform to Gelfand pairs. Some majors properties of this transform are examined. The localization operators related to the Stockwell transform in this framework are studied.

Spectral Barron spaces arising from quantum harmonic analysis

2026-03-06T22:05:25Z

In this paper, spectral Barron spaces are defined in the framework of quantum harmonic analysis. Their fundamental properties are studied. These include, among others, their completeness structure and some continuous embedding results. As an application, the existence and the uniqueness of the solution of a Schrödinger-type equation is proved.

On the sequential monotone closure of $CD_ω(K)$ spaces

2026-03-06T18:42:24Z

In this short note, we settle a problem posed by Wickstead in ~\cite{W:24}, arising from the study of the Riesz completion of spaces of regular operators between Banach lattices.

On amenability constants of Fourier algebras: new bounds and new examples

2026-03-06T17:08:39Z

Let $G$ be a locally compact group. If $G$ is finite then the amenability constant of its Fourier algebra, denoted by ${\rm AM}({\rm A}(G))$, admits an explicit formula [Johnson, JLMS 1994]; if $G$ is infinite then no such formula for ${\rm AM}({\rm A}(G))$ is known, although lower and upper bounds were established by Runde [PAMS 2006]. Using non-abelian Fourier analysis, we obtain a sharper upper bound for ${\rm AM}({\rm A}(G))$ when $G$ is discrete. Combining this with previous work of the first author [Choi, IMRN 2023], we exhibit new examples of discrete groups and compact groups where ${\rm AM}({\rm A}(G))$ can be calculated explicitly; previously this was only known for groups that are products of finite groups with ``degenerate'' cases. Our new examples also provide additional evidence to support the conjecture that Runde's lower bound for the amenability constant is in fact an equality.

$\mathcal{O}_α$-transformation and its uncertainty principles

2026-03-06T17:00:43Z

In this paper, we introduce a family of integral transforms, denoted by $\mathcal{O}_α$, and constructed via kernel fusion of the fractional Fourier transform (FRFT) with angle $α\notin π\mathbb{Z}$. We demonstrate that the $\mathcal{O}_α$-transformation constitutes a well-defined integral operator by establishing its basic operational properties. Besides, we survey various mathematical aspects of the uncertainty principles for the $\mathcal{O}_α$-transform, including Heisenberg's inequality, logarithmic uncertainty inequality, local uncertainty inequality, Hardy's inequality, Pitt's inequality, and Beurling-H{ö}rmander's theorem.

On semilinear Grushin--Schrödinger equation in $\mathbb{R}^N$

2026-03-06T15:55:19Z

We establish the existence of nontrivial nonnegative weak solutions to the following equation \begin{equation*} -Δ_γu + V(z)u = Q(z)f(u), \quad z\in \mathbb{R}^N, \end{equation*} where $Δ_γ$ denotes the so-called Grushin-type operator in $\mathbb{R}^N$. The potentials $V$ and $Q$ are assumed to be controlled below and above, respectively, by functions of type $(1+|z|)^a$, $a\in\mathbb{R}$. The main result is the embedded of the space $E_V^γ$ into the weighted Lebesgue space $L_Q^p(\mathbb{R}^N)$, under suitable conditions. Finally, we derive regularity results for the obtained weak solutions.

Local smoothing estimates for bilinear Fourier integral operators

2026-03-06T15:53:38Z

We formulate a local smoothing conjecture for bilinear Fourier integral operators in every dimension $d \ge 2,$ derived from the celebrated linear case due to Sogge, which we refer to as the \emph{bilinear smoothing conjecture}. We show that the linear local smoothing conjecture implies this bilinear version. As a consequence of our approach and due to the recent progress on the subject, we establish local smoothing estimates for Fourier integral operators in dimension $d=2,$ that is, on $\mathbb{R}^2_x \times \mathbb{R}_t$. Also, a partial progress is presented for the high-dimensional case $d\geq 3.$ In particular, our method allows us to deduce that the bilinear local smoothing conjecture holds for all odd dimensions $d$.

Mosco-convergence of Cheeger energies on varying spaces satisfying curvature dimension conditions

2026-03-06T14:39:23Z

We study the Mosco-convergence of Cheeger energies on Gromov-Hausdorff converging spaces satisfying different types of curvature dimension conditions. The case of functions of bounded variation is also considered. Applications to the continuity of Neumann eigenvalues are obtained. Our method, covering possibly infinite dimensional settings, is based on a Lagrangian approach and combines the stability properties of Wasserstein geodesics with the characterization of the nonsmooth calculus in duality with test plans.

Inequalities for Pairs of Measure Spaces and Applications

2026-03-06T13:02:06Z

We study a family of inequalities on pairs of measure spaces involving functions defined on product domains. Our main result establishes a Jensen-type inequality under a general product-measure framework, extending classical inequalities such as Hölder's and Minkowski's as special cases. The inequality admits sharp characterizations of equality and yields quantitative, variational, and probabilistic refinements under additional convexity assumptions. Several corollaries illustrate power-mean, entropy-type, and erasure-robust inequalities, as well as applications to convolution-type operators and weighted discrete models.

Continuity and equivariant dimension

2026-03-06T12:11:33Z

We study the local-triviality dimensions of actions on $C^*$-algebras, which are invariants developed for noncommutative Borsuk-Ulam theory. While finiteness of the local-triviality dimensions is known to guarantee freeness of an action, we show that free actions need not have finite weak local-triviality dimension. Moreover, the local-triviality dimensions of a continuous field may be greater than those of its individual fibers, and the dimensions may fail to vary continuously across the fibers. However, in certain circumstances upper semicontinuity of the weak local-triviality dimension is guaranteed. We examine these results and counterexamples with a focus on noncommutative tori and noncommutative spheres, both in terms of computation and theory.