Papers

One of the basic problems in mathematical learning theory is to identify a function $f: \Omega\to {\mathbb R}$ with certain specific properties which fits a given training data $\{(x_i, \xi_i)\in \Omega\times {\mathbb R}: i=1, \ldots, n\}$, in the sense that, $f(x_i) = \xi_i$ for $i=1, \ldots, n$, where $\Omega$ is a compact subset of ${\mathbb R}^d$ for some $d\in {\mathbb N}$ and $f$ is required to have certain specified characteristics. We address this problem when $f$ belongs to an arbitrary normed linear space ${X}$, and the evaluation maps $f\mapsto f(x_i)$ are replaced by maps of the form $ f\mapsto \varphi_i(f)$ on $X$, where $\varphi_1, \ldots, \varphi_n$ are continuous linear functionals on ${X}$. Using an inner product structure on ${\mathbb R}^n$, we shall device a method of least-squares for obtaining an approximate solution for the above problem and also identify a subspace of ${X}$ in which the least-square solution is unique, which in turn is shown to be equivalent to solving a matrix equation. In the context when the matrix under consideration is ill-conditioned, a regularized equation is devised, and order optimal error estimates are derived for the exact targeted data $\xi=(\xi_1, \ldots, \xi_n)$ and also when it is noisy, by choosing the regularization parameter appropriately.

We prove the strict log-concavity of the positive first eigenfunction \(-u\) of the \(2\)-Hessian equation and the strict $1/2$-convexity of the solution for the corresponding torsion problem in smooth bounded uniformly convex domains in $\mathbb{R}^{n}$. As applications, we establish the associated Brunn–Minkowski inequalities. We also show that this transformed-convexity phenomenon fails for \(3\)-Hessian equations by constructing, in dimension four, a smooth uniformly convex domain whose admissible zero-boundary solution has a nonconvex sublevel set.

We study data-driven prediction of coarse-grained dynamics in multiscale PDE systems. Adopting a closure-free operator-learning viewpoint, we apply a linear coarse-graining map and learn a surrogate evolution operator for the resolved field directly from filtered high-fidelity trajectories. Motivated by the Mori-Zwanzig formalism, we propose a spatiotemporal neural operator mapping a resolved history slab on $\Omega\times[-T_{\mathrm{in}},0]$ to a resolved future slab on $\Omega\times[0,T_{\mathrm{out}}]$. Spatial mixing uses Fourier convolution, while temporal mixing uses a causal kernel operator with position-attention weights on time lags. This causal temporal operator encodes finite-memory effects in the resolved dynamics while preserving the directionality of the history-to-future map. To improve rollout robustness and suppress nonconservative artifacts, we embed a flux-form inductive bias by parameterizing the windowed update in explicit divergence form. We also provide a data-driven guideline for selecting the memory length $T_{\mathrm{in}}$ via the decorrelation time of a closure-injection diagnostic computed from filtered trajectories. We validate on the coarse-grained viscous Burgers' equation, the Kuramoto-Sivashinsky equation, and two-dimensional turbulent flows, obtaining stable autoregressive rollouts with improved long-horizon accuracy and statistical fidelity.

We study the smooth topological equivalence between a nonautonomous linear system on the positive half-line and a quasilinear perturbation whose nonlinear part is not assumed to be globally bounded with respect to the state variable. The linear equation is assumed to admit a $(\mu,\nu)$–dichotomy, and the construction is carried out through the Green operator associated with this dichotomy. The usual global boundedness of the perturbation is replaced by locally uniform Green-integrability conditions along the relevant linear and nonlinear solutions. Under a smallness condition involving the global Lipschitz constant of the perturbation and the Green operator, we first construct Palmer-type maps which give a continuous topological equivalence on $\mathbb R^+$. We then impose Green-integrability conditions on the successive variational equations and a further first-order smallness condition which guarantees the invertibility of the derivative of the inverse map. Under these assumptions, the equivalence is of class $C^r$. We also provide a class of examples for which the perturbation is unbounded with respect to the space variable and all the hypotheses can be verified directly.

Each row of a transformer's attention matrix is a probability distribution over tokens, and in trained models most of that probability lands on a single sink token, usually the first. Standard tools for comparing attention rows (cosine similarity, Jensen–Shannon divergence, Shannon entropy) therefore hinge on a choice papers rarely report: keep the sink, or drop it and renormalize. This choice can reverse conclusions. On ten pretrained models from five families, 17–47% of verdicts about which of two heads is more similar flip with the convention, and the most prominent structure in a standard BERT head-clustering pipeline is an artifact of it. The reason is that one-number summaries mix two questions: how much attention the sink takes, and how the rest is divided among the content tokens. Treating rows as compositional data separates them exactly: the Aitchison distance splits orthogonally into a sink term and a content term, entropy splits by an exact identity, and the content distance is characterized by invariances the transformer itself possesses. The separation matters in practice: most measured entropy collapse during training is the sink growing, not attention sharpening (30% of the drop at 70M parameters, 95% at 1B, 79% at 1.4B), and pruning heads with the wrong channel can inflate perplexity more than a hundredfold. We map where each convention is safe, test a frozen out-of-sample predictor (one confirmation, one abstention, one failure), and release code regenerating every number.

This demo presents real-time AI-based uplink channel-estimation inference using data collected from a hardware-in-the-loop 5G platform. The data-collection setup integrates commercial RF signal generation, programmable channel emulation, an O-RAN Radio Unit, DU emulation, and a lightweight phase-aware convolutional neural network (CNN) that estimates the channel response directly from received DMRS signals. Unlike simulation-only evaluations, the hardware-derived dataset exposes the estimator to practical RF and system-level impairments, including calibration mismatches, synchronization imperfections, quantization effects, phase noise, and implementation-specific nonlinearities. During the demo, attendees will observe real-time CNN inference and channel reconstruction using captured hardware-generated DMRS observations and compare the proposed CNN against Least Squares (LS) and frequency-domain LMMSE baselines. The objective is to showcase a practical AI-native physical-layer inference pipeline that combines hardware-derived 5G data with real-time neural channel estimation for future 5G-Advanced and 6G systems.

We examine the configurationwise relation between the first Gribov horizon, defined by loss of positivity of the Faddeev-Popov operator, and Zwanziger's horizon function, which probes the inverse operator through background-dependent sources. The analysis focuses on whether the spectral directions associated with the onset of the Gribov horizon are necessarily accessible to the sources entering the horizon function, including situations in which the critical subspace is degenerate. This question is studied for radial SU(2) hedgehog backgrounds in three and four Euclidean dimensions, where angular symmetry constrains the source sector while the radial profile controls the ordering of Faddeev-Popov thresholds. Variational estimates and finite-volume calculations are used to characterize the threshold structure for smooth radial profiles. The regular-gauge BPST background is treated separately because of domain issues associated with zero-energy behavior and the horizon source in the full-space setting. The discussion is restricted to configurationwise spectral properties and does not address the statistical weighting of these backgrounds in the Yang-Mills functional integral.

In this study, we developed a unified geometric and relativistic framework for a new family of traversable wormholes constructed from a super-hyperbolic deformation of the classical catenoid, whose geometric embedding combines an explicit local curvature parameter $\kappa$, which secures a regular throat, with the Mittag-Leffler-based super-hyperboloid deformation function ${\mathfrak{S}C}_{\alpha} (v)$, where the deformation parameter $\alpha>1$ governs the radial onset and far-field growth of the embedding beyond the throat, continuously approaching the classical catenoidal geometry in the boundary limit $\alpha\rightarrow1^{+}$. Allowing a non-constant redshift $\Phi(r)$, we systematically evaluate closed analytical expressions for the shape function, curvature tensors, matter sources, and junction conditions, culminating in an exact thin-shell equation of motion and stability criterion. This work establishes a mathematically consistent and physically viable framework for the extension of wormhole theory, providing a new avenue for exploring exotic geometries within general relativity and identifying the conditions under which traversable configurations remain locally stable.

We investigate a notion of cospectrality for integer matrices that is parameterized by algebraic number fields. Given a number field and a symmetric integer matrix, we wonder when conjugating the integer matrix by an orthogonal matrix with entries in the given field can produce new integer matrices. Our results concern sufficient conditions for the associated notion of spectral determination, and we give constraints on the orthogonal matrices when the conditions are not applicable. The results use the discriminant of the characteristic polynomial and properties of Krylov subspaces. We leverage the theory to develop an algorithm to find all cospectral mates over a given (small) field. An implementation of the algorithm is made available.

The three gap theorem states that the points of the Kronecker sequence $\alpha,2\alpha,\ldots,N\alpha$, considered modulo one, divide the circle into intervals of at most three distinct lengths. In a two-dimensional nearest-neighbour analogue, Haynes and Marklof proved that the Kronecker sequence modulo an arbitrary unimodular lattice determines at most five distinct nearest-neighbour distances in the Euclidean norm, and that this bound is sharp. Dettmann subsequently constructed examples attaining five distinct distances for every $\ell_p$-norm, $1\leq p\leq\infty$. We prove the corresponding upper bound for every norm on $\mathbb R^2$: for every full-rank lattice $L$, every $\alpha\in\mathbb R^2$, and every $N\in\mathbb N$, the number of distinct nearest-neighbour distances is at most five. For every norm, this bound is attained for some $L$ and $\alpha$. For strictly convex norms, the proof extends the lattice-theoretic argument of Haynes and Marklof by replacing the Euclidean angular estimates with a cone lemma based on a proper Brass angular measure. The result for arbitrary norms is then obtained by a strictly convex perturbation and a limiting argument.

Let $X = \mathbb{P}^1 \smallsetminus \{0,1,\infty\}$ be the thrice-punctured over a ring of $S$-integers $\mathcal{O}_{K,S}$ in a number field~$K$. For any quotient $\pi_1^{\mathrm{mot}}(X,0) \twoheadrightarrow \Pi$ of Deligne–Goncharov's motivic fundamental group there is an associated Selmer scheme which parametrises $\Pi$-torsors with a mixed Tate motive structure. We give several descriptions of the motivic Selmer scheme which make it amenable to computations, using $\mathbb{G}_m$-equivariant cocycles of algebraic groups, Lie algebras, and complete Hopf algebras. We prove that the Selmer scheme is isomorphic to an affine space $\mathbb{A}^N_{\mathbb{Q}}$, and we construct coordinates realising this isomorphism. This is a key ingredient for making the motivic Chabauty–Kim method explicit in a general setting, without restrictions on the base field or the choice of fundamental group quotient.

We study the nonlinear stability of the elliptic equilibrium points $L_3$, $L_5$, and $L_6$ in the planar equilateral restricted four-body problem with unequal primary masses. We provide a systematic classification of nonlinear stability over the full two-parameter mass plane by combining high-order Birkhoff normal forms, numerical continuation, and singular reduction of resonant normal forms. In nonresonant regions, Arnold's theorem is applied when the quartic nondegeneracy condition holds, while sixth-order normalization resolves the degenerate cases. The $2$:$-1$ and $3$:$-1$ resonances are analyzed using the theorems of Alfriend, Markeev, and Meyer, including a classification of the $3$:$-1$ resonant dynamics. Singular reduction provides the corresponding reduced orbit spaces and reveals saddle-center bifurcations associated with the appearance and disappearance of periodic-orbit families. These results extend previous stability analyses beyond symmetric mass configurations and provide a geometric description of the resonant dynamics in the unequal-mass equilateral restricted four-body problem.

Graphically stable spaces $\overline{\mathcal{M}}_{g,G}$ parametrize marked nodal curves whose permitted collisions of markings are determined by a graph. We study intersection numbers of $\psi$-classes on $\overline{\mathcal{M}}_{g,G}$, as well as the classes $[M_{g,G}]$ in the Grothendieck ring of varieties. In both settings, we show that the geometry is governed by an underlying graphical structure, expressed through deletion-contraction relations. As consequences, we derive string and dilaton equations and express several families of $\psi$-class integrals in terms of the chromatic polynomial. We also express the Grothendieck class of $M_{0,G}$ over an arbitrary field in terms of the chromatic polynomial and identify various Euler characteristics with combinatorial quantities. Along the way, we obtain a new formula for Crapo's $\beta$-invariant of graphs. Finally, we extend these relations to genus one and, under a chromatic condition, to higher genus.

We study the implementable finite-batch particle algorithm for mean-field variational inference as a fully discrete stochastic approximation of the projected Wasserstein dynamics. The target potential is globally smooth but need not be strongly convex. The departure from contractivity is quantified by the curvature defect \[ \mathfrak d_\alpha(x,y) = \bigl[\alpha\|x-y\|^2- \langle\nabla V(x)-\nabla V(y),x-y\rangle\bigr]_+, \] which is the additive loss in the one-step Euler contraction estimate. We prove a non-asymptotic Wasserstein stability bound that separates initialization, product-empirical approximation, finite-batch drift error, time discretization, and the defects accumulated along the coupled trajectories. Under the uniform bound $\mathfrak d_\alpha\leq\beta$, the particle iterates remain within $O(\sqrt{\beta/\alpha})$ of any MFVI minimizer, up to explicit errors in the particle number, batch size, and step size. The proof uses a stationary comparison array whose population law is an MFVI minimizer but whose particle-level law is a random product empirical measure, and it controls the resulting projected-drift discrepancy explicitly. We also give coordinatewise defect estimates and structural conditions for dimension-independent projected-drift sensitivity, construct an arbitrary-dimensional smooth nonconvex benchmark with a closed-form MFVI minimizer, and explain why polynomially growing drifts require a modification of the untamed explicit scheme.

Submodular functions and their close relatives play a key role in combinatorial optimization, decision theory and potential theory. Part of their importance and usefulness stems from the connections with convex functions and polytopes. Here we explore connections between these functions and metric theory, with the bridge provided by diversities, a recently developed generalization of metric spaces to (finite) sets rather than just pairs. Both submodular functions and strongly submodular functions correspond to natural classes of diversities. Submodular diversities, as we define them here, are essentially non-decreasing, intersecting submodular functions which vanish on singletons. We prove new geometric embedding results for these diversities. In particular we show that submodular, strongly submodular, and XOS functions can be represented by the generalized circumradius, a set function in convex analysis equal to the amount a given convex body needs to be stretched to cover a set of points.

We provide an estimator for the perturbed utility route choice (PURC) model that works with data at the level of individual trips. The estimator is a nested fixed-point algorithm that combines an upper bias-corrected linear regression problem with a lower individual-level perturbed utility maximization problem. We establish the statistical properties of the microPURC estimator and confirm these results with an experiment using simulated data. Finally, we demonstrate the estimator in practice using a large real-world dataset.

We introduce a simple and performant approach for rapidly and accurately performing polyharmonic spline (PHS) interpolation using a combination of the fast multipole method (FMM) from computational electrostatics and the Vecchia approximation from the Gaussian process and sparse approximate inverse literatures. Using basic properties about Hadamard products and low-rank matrices, we demonstrate that an FMM with two kernels, the logarithmic and distance kernels, results in fast PHS interpolation for all orders. Furthermore, we demonstrate the exceptional performance of sparse inverse approximation methods with the Matérn covariance model for preconditioning. Combined with careful management of disallowed subspaces, we describe a procedure for obtaining prediction weights using preconditioned conjugate gradient that converges in less than $15$ iterations, even for problem sizes with over one million points. As a result, thin-plate spline interpolation—a particularly popular method that does not require parameter tuning—that matches the fully dense $\mathcal{O}(n^3)$ computation in accuracy can be done at the cost of approximately $50-60$ FMMs. A high-performance software library for odd-order PHS interpolation in two dimensions is made available as a companion to this work.

Given a configurational partition function, in this work we investigate the inverse problem of reconstructing the one-dimensional potential. For the case of the Gaussian partition function, the configurational density of states (CDOS) is univocally determined, being the harmonic potential uniquely recovered due to its symmetric single–well feature. When multiples inverse branches are considered, the uniqueness of the potential is broken and more information is needed in order to obtain a unique potential. The branch topology imposes alternating orientations, thus allowing distinct spatial realizations. Finally, the Gaussian partition function only determines the CDOS but not the potential, manifesting in this way the precise scope and limitations of the inverse characterization of the Gaussian configurational partition function.

We prove that, for arbitrarily small positive $\varepsilon$, there is a finite planar point set $P$ such that every $\varepsilon$-net for the range space induced on $P$ by straight lines has cardinality $\Omega\bigl(1/\varepsilon \cdot \log(1/\varepsilon)\bigr)$. This matches the classical upper bound for range spaces with bounded VC-dimension due to Haussler and Welzl and confirms a prediction of Alon.

Circular photon orbits play a pivotal role in both gravitational theories and astronomical observations. However, the properties of circular photon orbits in naked singularity spacetimes remain insufficiently explored and deserve in-depth investigation. The present work is dedicated to a comprehensive study of the features of circular photon orbits in naked singularity spacetimes. Notably, a geometric approach is employed to investigate these orbits, in which the framework of optical geometry together with its intrinsic curvatures plays a crucial role. We analyze the existence of circular photon orbits through the intrinsic geodesic curvature, and then examine the numbers of stable and unstable circular orbits, as well as the topological invariant associated with circular photon orbits. By examining various classes of naked singularity spacetimes, we obtain a general conclusion regarding circular photon orbits that holds for the most of spherically symmetric naked singularity spacetimes: the total number of circular photon orbits is an even integer ($\textit{N} = \text{2}\textit{k}$), consisting of $\textit{k}$ stable orbits and $\textit{k}$ unstable orbits in equal proportion. A mathematical proof of this conclusion is also provided in the present work. Furthermore, a comparison with other categories of spacetimes reveals that the conclusions regarding the number of circular photon orbits in naked singularity spacetimes agree with those obtained for compact object spacetimes without naked singularities, indicating that the number of circular photon orbits is primarily governed by the presence of event horizons rather than spacetime singularities. Keywords: Circular Photon Orbit, Naked Singularity Spacetime, Optical Geometry, Geometric Curvatures

The Tsallis $q$-sum is one of the fundamental nonlinear composition laws of nonextensive statistical mechanics. Although it has been extensively investigated in continuous settings, its implications for finite mathematical structures remain less explored. Here we investigate this question by replacing ordinary additive laws with the Tsallis $q$-sum. We first introduce an axiomatic $q$-cardinality of finite sets satisfying a nonlinear additivity principle. Existence and uniqueness are established, leading to an explicit expression that continuously recovers the classical cardinality as $q\to1$. We then propose a nonlinear matrix composition induced by the same deformation and establish its principal algebraic properties, including associativity, the neutral element, and a characterization of its noncommutativity in terms of the ordinary matrix commutator. An illustrative example over $M_2(\mathbb Z_6)$ shows how the deformation can modify the center of a finite matrix algebra when $1-q$ is a zero divisor. These results indicate that the Tsallis $q$-sum provides a natural mechanism for constructing nonlinear finite mathematical structures and connects nonextensive statistical mechanics with finite algebra and nonlinear mathematical physics.

We prove the optimal convergence rate $O(\varepsilon)$ for periodic homogenization of convex Hamilton-Jacobi equations arising from infinite systems of indistinguishable particles on the torus, under the assumption that the initial data depend only on the mean configuration. This extends the finite-dimensional result [1], which is based on the large-time behavior of the Lagrangian action metric and a curve-surgery argument. Here, these tools cannot be applied directly because minimizing curves live in an infinite-dimensional Hilbert space, where local compactness and finite-dimensional topology are unavailable. We overcome this difficulty by cutting the finite-dimensional mean-time projection of a minimizing curve and gluing the lifted pieces in the Hilbert space using the compact quotient induced by periodicity and rearrangement invariance. We conclude with an example showing that this rate is sharp.

Every bounded linear operator $T$ on a complex Banach space $E$ is known to extend to a lattice homomorphism $\overline{T}$ that acts on the so-called complex free Banach lattice over $E$. We show the following three results about the spectrum $\sigma(\overline{T})$ of $\overline{T}$: (i) $\sigma(\overline{T})$ always contains the spectrum $\sigma(T)$; this answers a recent question of de Hevia and Tradacete. (ii) It can happen that $\sigma(T)$ is a singleton while $\sigma(\overline{T})$ is the entire unit circle; this shows that $\sigma(\overline{T})$ is not the closure of the cyclic hull of $\sigma(T)$ in general. (iii) Finally, we give a full characterization of $\sigma(\overline{T})$ in terms of the spectral properties of $T$. Some of our arguments also give new results about the spectral properties of general lattice homomorphisms. As a main tool we make extensive use of the Banach lattice functional calculus for continuous positively homogeneous functions.

We study the capacity of a memory-based quantum repeater in entanglement swapping between two quantum links with either single or multiple memories, which we refer to as the end-to-end (E2E) entanglement throughput, subject to a constraint on the minimum fidelity. In order to approximate the E2E entanglement throughput, we adopt queueing models, where quantum links can have different characteristics: memory capacities, entanglement attempt rates and success probabilities, as well as classical communication latencies. We develop a model for estimating E2E entanglement fidelity, while taking into account the heterogeneous dephasing and depolarizing dynamics of quantum memories and Bell-state measurements in entanglement swapping as well as classical communication delays and noises. Finally, with the help of our models for approximating the E2E entanglement throughput and fidelity, we use the maximum waiting times of entanglements in quantum memories at the repeater as optimization variables to maximize the E2E entanglement throughput while ensuring required minimum E2E fidelity.

It has been recently shown that the notoriously divergent $L ^2 $ metric on the moduli space of hyperbolic monopoles can be made finite by the introduction of a modified gauge fixing condition \cite{franchetti:2024}. In this paper we compute this modified $L ^2 $ metric for the mass $\tfrac{1}{2}$ 2-monopoles. The resulting metric is actually a complex non-degenerate bilinear form, which restricts to a Riemannian metric on the 4-dimensional subspace of inversion symmetric 2-monopoles. Remarkably, we have been able to compute the $L ^2 $ form explicitly in terms of elementary functions and elliptic integrals. Its asymptotic form matches what was expected on the basis of previous results in the literature.

Erickson and Jones classified the nontrivial homometry classes of five-point binary bracelets into seven families, Types A-G, and conjectured that each family is invariant under multiplication by units modulo sign. We prove this conjecture and show that the induced actions are diagonal for Type A, multiplicative for Types B-E, affine for Type F, and regular through $U_{20}/\{\pm1\}\cong C_4$ for Type G. We derive closed arithmetic formulas for the numbers of unit orbits and classes fixed by every unit, and determine the orbit-size distributions in all seven types, including explicit stabilizer descriptions for Types A and F.

The Performance estimation problem (PEP) approach reformulates finding the exact worst-case performance of an algorithm as the solution to an optimization problem. Tractable formulation of the problem requires necessary and sufficient interpolation conditions. We present the interpolation conditions for convex–concave functions, bilinear functions, and composite functions with bilinear coupling. We also construct PEP for first-order methods for composite saddle-point problem.

Public Transport (PT) lines are traditionally designed to optimize performance under nominal traffic conditions. In practice, operating conditions frequently deviate from nominal ones, leading to substantial performance deterioration. Existing adaptation mechanisms typically rely on reactive interventions, such as stop-skipping, which are insufficient under large or recurrent traffic fluctuations. In such contexts, incremental adjustments may not suffice. This paper evaluates the potential of deeper structural redesigns to preserve performance. We propose a method to proactively redesign appropriate parts of PT networks under high traffic fluctuations that would otherwise deteriorate operator and user performance. We adopt a predict-then-optimize paradigm in which PT lines are reconfigured based on traffic forecasts using the Non-dominated Sorting Genetic Algorithm III (NSGA-III). To ensure operational feasibility and avoid excessive structural changes, we enforce high Jaccard edge overlap between the original and redesigned networks. To assess prediction inaccuracies, we construct a statistical model of errors from a well-established deep learning predictor, the Diffusion Convolutional Recurrent Neural Network, trained on real-world data. Computational results on Mandl's benchmark and the large-scale Beijing network show that controlled PT line redesign yields substantial user-centric performance gains and operational cost reductions under high traffic fluctuations while limiting topological changes. On the Beijing network under high variability, average travel time improves by up to 25.8% while preserving over 85% line overlap. Unlike stop-skipping baselines, which break connectivity for many OD pairs, our redesign preserves full OD connectivity. These results support a shift from static planning toward continuous and adaptive PT network design.

The Cox model and the Cox regression coefficient are widely used in applied work, although the corresponding statistical theory is principally developed in the continuous case where there can be no tied failure times. In this work, we derive the efficiency bound of the Cox regression coefficient in the Cox model, without imposing continuity or discreteness assumptions on the failure distribution. Then, under a technical assumption that the population failure mass points belong to an unknown finite set, we show that the “exact” estimator in \citet{cox1972regression} is asymptotically efficient. Next, we propose an asymptotically equivalent estimator which solves the efficient score which has lower computational complexity than the exact Cox score. The development employs recently introduced martingale theory, and throughout examples of the general theory are given for both the absolutely continuous case and the discrete case.

We introduce Probabilistic Ball Mapper, a formulation of Ball Mapper in which each data point is assigned a probability distribution supported only on the metric balls that contain it. This assignment defines both a partition subordinate to the Ball Mapper cover and a Markov kernel from the finite data space to the cover. We study two assignment schemes: a uniform-on-support rule and a localized radial-basis rule that incorporates distance to landmarks while preserving the underlying cover. Pushing the empirical data distribution through the kernel produces a probability distribution over vertices. Drawing twice, conditionally and independently, from each pointwise distribution produces a soft-overlap matrix. This matrix is symmetric, nonnegative, positive semidefinite, and has the vertex distribution as both marginals. It therefore provides a mass-normalized refinement of classical Ball Mapper overlap rather than another unnormalized edge count. For graphs constructed on a common cover, the vertex and overlap distributions can be compared directly. For independently fitted covers, we formulate Wasserstein and fused Gromov–Wasserstein-type discrepancies that account for vertex mass, landmark geometry when a common ambient metric is available, and intrinsic graph relations. For a fixed cover, we derive explicit perturbation bounds controlled by the sensitivity of the assignment rule, the magnitude of the data perturbation, and the data mass near cover boundaries. When the cover is recomputed, landmark motion creates an additional source of variation, for which we state a transport-based stability principle rather than an unconditional theorem. The resulting framework turns Ball Mapper into a probability-valued representation suitable for quantitative comparison while retaining its geometric interpretability and computational simplicity.