Combinatorics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
Monday 11 May
Let $k\geq 2$. We show that, for a sufficiently small $\varepsilon>0$, any sufficiently large $n$-vertex Hamiltonian graph of minimum degree at least $n^{1-\varepsilon}$ contains a $2$-factor consisting of exactly $k$ cycles. This is the first minimum-degree condition which is polynomially smaller than linear. Our methods yield an analogous result when the host graph is not required to contain a Hamilton cycle, but only a $2$-factor consisting of at most $k$ cycles; this answers a question of Bucić, Jahn, Pokrovskiy and Sudakov.
Statistics Theory
A Data-Consistent Approach to Ensemble Filtering
Ensemble data assimilation is challenging when the ensemble size is substantially smaller than the state and observation dimensions, leaving empirical covariance information rank deficient and sensitive to finite sampling. We introduce the Quantity-of-Interest Principal Component Analysis Ensemble Data-Consistent Filter (QPCA-EnDCF), a deterministic filtering method motivated by data-consistent inversion that regularizes the update in the whitened forecast–observation mismatch space. The method restricts the correction to a rank-$\kappa$ empirical spectral subspace before mapping it to state space through an empirical EnDCF gain. We establish concentration and spectral-projector perturbation results and derive a bias–variance decomposition showing that truncation bias is governed by the discarded component of the population mean residual, while finite-ensemble variability depends on estimation of the retained correction. QPCA-EnDCF avoids the analysis-mean variance contribution from perturbed observations in stochastic EnKF updates, and Lorenz–96 experiments show improved calibration and lower mean-squared error relative to the tested stochastic and deterministic controls, primarily through lower across-realization variance despite increased bias. An idealized inlet shallow-water experiment further shows that QPCA-EnDCF can be implemented in a high-dimensional, severely rank-deficient nonlinear model and remains numerically viable through the prescribed shelf-crossing endpoints.
Read the full abstract
Probability
Flocking with Multiple Types: Competition, Fluid Limits and Traveling Waves
We study a class of interacting particle systems on $\mathbb{R}$ with two types. Particles evolve by independent jumps sampled from a fixed distribution, with type-dependent jump rates $v_+$, $v_-$ and stochastic type switching driven by non-local order-based interactions. The switching rates depend on the empirical distribution through the proportion of opposite-type particles located ahead, leading to a nonlinear and discontinuous dependence on the empirical measure outside the standard Lipschitz McKean-Vlasov framework. Our first main result is a law of large numbers for the empirical measure process: we prove convergence, along subsequences, to a deterministic measure-valued process characterized by a McKean-Vlasov equation. The proof combines tightness in Wasserstein space with a martingale characterization of limit points. A uniqueness argument based on a Kolmogorov-Smirnov-type distance adapted to the ordering structure yields convergence of the full empirical measure sequence and, in turn, propagation of chaos on finite time intervals. We then study the long-time behavior of the limiting dynamics. Because the system has persistent drift, invariant distributions do not arise; instead, we analyze traveling waves, corresponding to stationary profiles in a moving frame. For exponential jump distributions, the associated non-local integro-differential system admits a local description. In the regime $v_+>v_-=0$, this further reduces to a coupled system of non-linear ODEs, allowing a phase-plane analysis that yields a traveling wave as a heteroclinic orbit connecting two equilibria. We also identify the wave speed and mass partition, and derive tail asymptotics by spectral analysis of the linearized system.
Read the full abstract
Differential Geometry
Convergence of Lorentzian spaces and curvature bounds for generalized cones
The goal of this article is twofold. We introduce a notion of convergence for Lorentzian pre-length spaces, $\ell$-convergence, that extends previous convergence notions in this context. We show that timelike curvature and timelike curvature-dimension bounds are stable under (measured) $\ell$-convergence. Then, we show that $\ell$-convergence is well adapted for generalized Lorentzian cones: a sequence of generalized cones $-I_i\times_{f_i}X_i$ converges in $\ell$ sense if the base $I_i$ and the fiber $X_i$ converge in GH sense and the functions $f_i$ converge uniformly. We use this to show sharp timelike curvature and timelike curvature-dimension bounds for such cones. Finally, we obtain a pre-compactness theorem for $\ell$-convergence in the class of smooth generalized cones that have a uniform lower bound on the full Ricci (or Riemann) curvature tensor.
Read the full abstract
Optimization and Control
A Unified Approach for Computing Wasserstein Barycenters of Discrete and Continuous Measures
Computing the unregularized Wasserstein barycenter for measure-valued data is a challenging optimization task. Recent algorithms have been tailored to either discrete measures as point clouds or continuous measures discretized on regular grids. In this work, we propose a primal mirror descent algorithm for computing the exact Wasserstein barycenter in the Fisher-Rao geometry. Our algorithm is a unified approach that is flexible enough to simultaneously cover discrete and absolutely continuous input measures, with convergence guarantees in both settings. In particular, when all input measures are discrete, our algorithm, initialized from any probability density, solves a sequence of semi-discrete optimal transport subproblems and produces absolutely continuous iterates that converge to the discrete barycenter. We use synthetic and real data examples to demonstrate the promising result in terms of accuracy and computational cost.
Read the full abstract
Probability
Uniform sampling of multitype continuous-time Bienaymé-Galton-Watson trees
We study the genealogy of a sample of $k$ individuals taken uniformly without replacement from a continuous-time multitype Bienaymé–Galton–Watson process at fixed times. Our results are quite general, requiring only that the process be non-simple and conservative, and that every type has a positive probability to “eventually lead to” all other types within the population. The corresponding single-type case has recently been studied by Johnston (2019), Harris, Johnston, and Roberts (2020), and Harris, Johnston, and Pardo (2024). Our approach is based on a $k$-spine decomposition and a suitable change of measure under which the distinguished spines form a uniform sample at time $T$, while the population size is subject to $k$-size biasing and exponential discounting. This construction preserves a branching Markov property and yields an explicit description of the genealogical tree at fixed times. In particular, we characterise spine splitting times, offspring distributions, and type-dependent ancestral structures, revealing rich interactions between types that are absent in the single-type setting. The present results form the basis of a forthcoming series of papers in which limiting genealogical behaviour is analysed under various asymptotic regimes and more general sampling schemes by the authors, see Angtuncio et al. (2026b), (2026c) and (2026d).
Read the full abstract
Analysis of PDEs
Global well-posedness for the Hele-Shaw problem with point injection
We study the two-dimensional Hele-Shaw problem with point injection for star-shaped domains. We reduce the system to a nonlocal parabolic equation of the interface, and for arbitrary Lipschitz initial interface away from the source, we prove global well-posedness of the interface equation in a strong sense. We also introduce a viscosity-solution framework for the interface equation and relate it to the classical viscosity theory for the Hele-Shaw problem.
Read the full abstract
Combinatorics
Averaged Extensions of Golomb's Triangular Recursion: Critical Invariance and Supercritical Constraints
For an integer $m\ge 1$ and a parameter $\alpha>0$, consider the nested recursion $$ Q_{\alpha,m}(n)=1+\left\lfloor \frac{\alpha}{m}\sum_{j=1}^m Q_{\alpha,m}\!\left(n-Q_{\alpha,m}(n-j)\right)\right\rfloor, \qquad n>m, $$ with $Q_{\alpha,m}(1)=\cdots=Q_{\alpha,m}(m)=1$. For $m=1$ and $\alpha=1$, this is Golomb's non-homogeneous triangular recursion. We prove that its canonical triangular solution is preserved, up to an initial index shift, by every finite arithmetic averaging length. More generally, the same exact solution is generated by any aggregator satisfying a local floor-lock condition. This class includes all power means of finite order, including the harmonic and geometric means, as well as the minimum and positively weighted quasi-arithmetic means. For $m\ge 2$, the maximum lies outside this class but has a different explicit block law. Consequently, every value $k\ge 2$ occurs exactly $k$ times in the floor-admissible class, and $$ Q_{1,m}(n)=\left\lfloor\frac{1+\sqrt{1+8(n-m)}}{2}\right\rfloor \sim\sqrt{2n}. $$ For $0<\alpha<1$, the sequence is identically one. Near criticality, with $\alpha=1+\delta$ and $0<\delta<(2m-1)^{-1}$, we determine the exact first departure time from the critical orbit, of order $\delta^{-2}$. We also prove a finite-step breakdown criterion for large $\alpha$ and a conditional slope theorem: any globally defined solution with a limiting density in $(0,1)$ must have slope $1-\alpha^{-1}$. Exact-arithmetic computations support, but do not prove, a supercritical linear-growth regime.
Read the full abstract
Differential Geometry
Robin nullity in mode $|k|=1$ and asymptotic radius of the critical spherical catenoid
Medvedev proved that the critical spherical catenoid $Sigma_a$, the rotationally symmetric free boundary minimal annulus in a geodesic ball $B^3(r(a))$ in $H^3$ from the family of Mori and do Carmo-Dajczer, has Morse index at least 4, and conjectured equality. For each $a>1/2$ we establish three analytic results about $Sigma_a$. (I) Robin nullity and index in mode $|k|=1$. The Robin nullity of the Jacobi operator $L_{\Sigma_a}=\Delta_g+(|II|^2-2)$ in angular Fourier mode $|k|=1$ equals $2$, with kernel spanned by the Killing–Jacobi fields associated to the rotations $L_{12},L_{13}\in\mathfrak{so}(3,1)$ that fix the geodesic axis of $\Sigma_a$ and send $\partial B^3(r(a))$ to itself. The radial profile admits the closed form $f_*(s)=\partial_s\Phi_a^0(s,0)=\frac{d}{ds}[A(s)\cosh\varphi(s)]=\sinh r(s)\cdot r'(s)$, where $r(s)$ is the geodesic distance from $p_0=(1,0,0,0)$. By Sturm–Liouville theory, the Robin Morse index of $\Sigma_a$ in mode $|k|=1$ also equals $2$, refining the lower bound of Medvedev. (II) Asymptotic radius. The boundary radius satisfies $r(a)=\tfrac{3}{2}\log a+d_\infty+o(1)$ as $a\to\infty$, with $d_\infty=\log[\sqrt{2}\,\Gamma(1/4)^2/\pi^{3/2}]=\log[2\sqrt{2\pi}/\Gamma(3/4)^2]$. The closed form for $d_\infty$ follows from a Beta-function evaluation of $I_\infty=\int_0^{\infty}\cosh(2t)^{-3/2}\,dt$. (III) Degenerate limit. As $a\to(1/2)^+$, $r(a)=c_*\sqrt{a-1/2}\,(1+o(1))$ with $c_*=\sigma_*\cosh\sigma_*$, where $\sigma_*$ is the unique positive fixed point of $\sigma=\coth\sigma$. The proof of (I) follows the mode-by-mode strategy of Devyver for the Euclidean critical catenoid, with $\mathfrak{so}(3,1)$ replacing $\mathfrak{so}(3)$, supplemented by the closed-form identification $f_*=\partial_s\Phi^0$ specific to the hyperbolic ambient. The proof of (II) is a Laplace-type asymptotic analysis of the implicit free boundary condition.
Read the full abstract
Combinatorics
Urschel Nodal Domains via Perturbation Theory
We prove several types of Courant nodal domain theorems for generalized Laplacians on graphs, based on an invariant introduced by Urschel, which we call the "Urschel number", denoted ${\rm UN}({\bf f})$, of an eigenvector ${\bf f}$. We refine Urschel's invariant, and use perturbation techniques to obtain some new results. First, we show the existence of mutually orthogonal eigenvectors, such that if the $k$-th eigenvalue has multiplicity $m$, then for $0\le j\le m-1$, ${\rm UN}({\bf f}_{k+j})\le k+\min(j,(m-1)-j)$. Second, for a simple $k$-th eigenvalue, we classify the zeroes of ${\bf f}_k$ as either "shallow or "deep"; we obtain a number of results that say, roughly speaking, the more shallow vertices ${\bf f}_k$ has, the more control we have over our new invariants based on Urschel's. Our new invariants of an eigenvector, ${\bf f}_k$, are a sequence of integers whose minimum value is ${\rm UN}({\bf f}_k)$ and whose maximum, denoted ${\rm UN}_{\max{}}({\bf f}_k)$, is the maximum number of nodal domains of any possible positive/negative signing or "charge" of the zeroes of ${\bf f}_k$. An example of our second type of result is that if ${\bf f}_k$ has no deep vertices, then ${\rm UN}_{\max{}}({\bf f}_k)\le k$. We provide a number of examples to illustrate our main results, and how they differ from the situation in analysis. We also describe a minor improvement of the Gladwell-Zhu theorem for an orthonormal eigenbasis in the presence of eigenvalues of sufficient multiplicity.
Read the full abstract
Statistics Theory
Efficient Robust Constrained Signal Detection via Kolmogorov Width Approximations
Robust statistical inference often faces a severe computational-statistical gap when dealing with complex parameter spaces. We investigate minimax signal detection in the Gaussian sequence model under strong $\epsilon$-contamination, where the signal belongs to a general prior constraint $K$. Existing optimal tests require computing the exact Kolmogorov $k$-width of $K$, a computationally intractable task for general non-trivial sets. We bridge this gap by proposing a polynomial-time testing framework that universally applies to balanced, type-2, and exactly 2-convex constraints. By leveraging a semidefinite programming relaxation and a modified ellipsoid method equipped with an approximate subgradient oracle, we efficiently approximate the Kolmogorov widths. Remarkably, our unconditional efficient algorithm achieves a robust detection boundary that matches existing upper bounds up to a mere polylogarithmic factor. This establishes a computationally tractable testing solution for a broad class of structured signals without requiring prior knowledge of their exact geometric complexity.
Read the full abstract
Algebraic Topology
A Stable Distance Persistence Homology for Dynamic Bayesian Network Clustering
Dynamic Bayesian networks (DBNs) are a widely used framework for modeling systems whose probabilistic structure evolves over time. Standard inference methods focus on local conditional distributions and can miss larger-scale patterns in how dependencies between variables organize and change over time. We introduce a topological approach to this problem. To each DBN we associate a time-varying graph, called a Dynamic Bayesian Graph (DBG), by assigning to each edge a strength that measures variation in its conditional dependence across parent configurations, and retaining edges whose strength exceeds a chosen threshold. We show that this construction fits within the dynamic graph framework of Kim and Mémoli, enabling the use of tools from topological data analysis. Applying persistent homology to a DBG produces a barcode, which records the merging and disappearance of connected groups of strongly dependent variables over time. We prove that this barcode is stable: small perturbations in the conditional probability tables of the DBN lead to small changes in the resulting barcode. This yields a principled and noise-resistant summary of how dependency structure evolves in a dynamic Bayesian network.
Read the full abstract
Representation Theory
Balanced subsets in root systems
Algebraic Geometry
Automorphism Groups of Reductive and Root Monoids
We describe the automorphism groups of reductive monoids and of root monoids with active groups of invertible elements.
Algebraic Topology · Cross-listed from eess.SY
Multi-Agent System Identification with Nonlinear Sheaf Diffusion
Local interaction laws governing multi-agent systems can be difficult to recover from trajectory data, even when the dynamics are observed faithfully. In systems governed by a nonlinear sheaf Laplacian – a generalization of the graph Laplacian accommodating heterogeneous state spaces and asymmetric communication channels – the coordination law is encoded by edge potential functions whose gradients produce the inter-agent forces. Because trajectory observations record node-state evolution, they expose only the aggregate effect of the edge forces at each node: distinct interaction laws that agree at the node level are indistinguishable from trajectory data alone. We show that the fundamental obstruction to recovery is topological, measured by sheaf cohomology, and that unique recovery from an unconstrained function class is possible if and only if this cohomology vanishes. When the obstruction is nontrivial, we show that recovery within a finite-dimensional parameterized class is possible precisely when a data-dependent information matrix is positive definite. Experiments validate the theory and illustrate that accurate trajectory reproduction need not certify recovery of the underlying interaction law.
Read the full abstract
Functional Analysis
Functional Models of Abelian Locally von Neumann Algebras and Direct Integrals of Locally Hilbert Spaces
We obtain a functional model for an arbitrary Abelian locally von Neumann algebra acting on a representing locally Hilbert space under the assumption that the index directed set is countable, in terms of locally essentially bounded functions on strictly inductive systems of measure spaces, which can be viewed as the reduction theory of this kind of operator algebras. Then, we single out the concept of a direct integral of locally Hilbert spaces and the concepts of locally decomposable and locally diagonlisable operators and we show that these form locally von Neumann algebras that are commutant one to each other. Finally, we show that any Abelian locally von Neumann algebra, which acts on separable representing locally Hilbert spaces and such that the index set is a sequentially finite directed set, is spatially isomorphic with the Abelian locally von Neumann algebra of all locally diagonlisable operators on a certain direct integral of locally Hilbert spaces with respect to a certain strictly inductive system of locally finite measure spaces on standard Borel spaces.
Read the full abstract
Algebraic Geometry
Evaluation codes from linear systems of conics
The Datta-Johnsen code is an evaluation code where the linear combinations of elementary symmetric polynomials are evaluated on the set of all points with pairwise distinct coordinates in an affine space of dimension $\ge 2$ over a finite field $\mathbb{F}_q$. A generalization is obtained by taking a low dimensional linear system of symmetric polynomials. The odd characteristic case was the subject of a recent paper. Here, the even characteristic case is investigated.
Read the full abstract
Numerical Analysis · Cross-listed from cs.LG
Muon is Not That Special: Random or Inverted Spectra Work Just as Well
The recent empirical success of the Muon optimizer has renewed interest in non-Euclidean optimization, typically justified by similarities with second-order methods, and linear minimization oracle (LMO) theory. In this paper, we challenge this geometric narrative through three contributions, demonstrating that precise geometric structure is not the key factor affecting optimization performance. First, we introduce Freon, a family of optimizers based on Schatten (quasi-)norms, powered by a novel, provably optimal QDWH-based iterative approximation. Freon naturally interpolates between SGD and Muon, while smoothly extrapolating into the quasi-norm regime. Empirically, the best-performing Schatten parameters for GPT-2 lie strictly within the quasi-norm regime, and thus cannot be represented by any unitarily invariant LMO. Second, noting that Freon performs well across a wide range of exponents, we introduce Kaon, an absurd optimizer that replaces singular values with random noise. Despite lacking any coherent geometric structure, Kaon matches Muon's performance and retains classical convergence guarantees, proving that strict adherence to a precise geometry is practically irrelevant. Third, having shown that geometry is not the primary driver of performance, we demonstrate it is instead controlled by two local quantities: alignment and descent potential. Ultimately, each optimizer must tune its step size around these two quantities. While their dynamics are difficult to predict a-priori, evaluating them within a stochastic random feature model yields a precise insight: Muon succeeds not by tracking an ideal global geometry, but by guaranteeing step-size optimality.
Read the full abstract
Representation Theory · Cross-listed from cs.LG
Oversmoothing as Representation Degeneracy in Neural Sheaf Diffusion
Neural Sheaf Diffusion (NSD) learns a task-adapted sheaf geometry for node classification. We connect the study of these learned geometries with Geometric Invariant Theory (GIT) to analyse their representation-theoretic structure and develop mechanisms for geometric control during training. Identifying cellular sheaves with representations of the incidence quiver connects direct-sum decompositions to the harmonic space and motivates the study of trivial or low-complexity summands as potential degeneration modes. We make this perspective concrete by focusing on the trivial representation: on an embedded copy, the sheaf Dirichlet energy reduces to weighted scalar graph energy. The GIT stability framework provides a criterion for excluding such copies within the semistable class. Using its moment-map characterisation of polystability, we construct differentiable surrogates that encourage the corresponding balance conditions. We identify an equal-stalk obstruction to assigning a destabilising weight to the trivial representation and show how nonuniform stalk dimensions remove it. Controlled energy-minimisation experiments reproduce trivial-component formation, while native feature-conditioned NSD experiments on heterophilic graphs show how moment-map regularisation changes the balance and Dirichlet-energy trajectories of learned geometries during training, alongside their classification behaviour.
Read the full abstract
Differential Geometry
Explicit Laplace Spectra of Homogeneous Principal Bundles
We present a unified representation-theoretic method to compute the Laplace-Beltrami spectrum on homogeneous principal bundles. For this setting, we introduce a multi-parameter family of metric deformations called generalized canonical variations. Building upon the geometric realization of such fibrations as naturally reductive spaces, we establish a simplified spectral branching criterion. We apply this method to derive the full spectra (yielding all eigenvalues and multiplicities) for several prominent geometric families. Specifically, we compute the full spectra for the entire classical series of homogeneous 3-$(\alpha,\delta)$-Sasaki manifolds (Types A, B, C, and D) and for real and complex Stiefel manifolds over Grassmannians. These explicit formulas provide the analytical data to investigate related problems in geometric analysis. As an application, we classify the scalar stability of these spaces under Perelman's $\nu$-entropy and, for the 3-$(\alpha,\delta)$-Sasaki manifolds, determine the exact thresholds for Yamabe bifurcations.
Read the full abstract
Numerical Analysis
An energy-decreasing algorithm for the finite element approximation of ferronematic equilibrium states and its numerical realization
We develop an energy-decreasing algorithm for the finite element approximation of two-dimensional ferronematic equilibrium states. The problem is formulated as the minimization of the harmonic energy of two two-dimensional vector fields, both with prescribed length, together with an additional nonlinear relation on the orientation of the two vectors. The finite element setting is based on piecewise continuous finite elements on a weakly acute triangulation. The numerical realization of the energy-decreasing algorithm employs a decomposition-coordination framework and a Uzawa-like iteration. Numerical experiments are presented to illustrate the computational performances of the algorithm.
Read the full abstract
Algebraic Geometry
Functoriality of logarithmic Hochschild homology of log smooth pairs
The construction of a satisfactory dg category of logarithmic coherent sheaves remains a central open problem in logarithmic geometry. In this paper, we propose an alternative correspondence-theoretic approach based on logarithmic Fourier–Mukai transforms. For smooth proper log pairs, we introduce strong log Fourier–Mukai kernels supported on canonical blow-up compactifications and prove that logarithmic Hochschild homology is functorial with respect to the induced transforms. Unlike the classical setting, logarithmic correspondences do not naturally live on ordinary products, and the standard adjunction formalism fails because of blow-up discrepancies. We overcome these difficulties by constructing explicit unit- and counit-type morphisms that provide the necessary adjunction data without requiring an ambient dg category of logarithmic sheaves. As applications, we construct a dg bicategory of logarithmic correspondences in which logarithmic Hochschild homology and cohomology become categorical invariants. We also define logarithmic Chern characters and a logarithmic Euler pairing compatible with the logarithmic Fourier–Mukai formalism.
Read the full abstract
Algebraic Geometry
Theta functions for singular curves
Let $X$ be an irreducible singular Riemann surface, with desingularisation $\widetilde X$. The generalised Jacobian $J(X)$ of $X$ fibers over the Jacobian $J(\widetilde{X})$ of $\widetilde X$, and there is an Abel map $A$ of $\widetilde X$ to $J(X)$, lifting the Abel map to $J(\widetilde X)$. We build a theta function on a compactification of the generalised Jacobian $J(X)$ (giving a section of a suitable positive line bundle). The translation action on $J(X)$ then yields all line bundles of that degree, and the translates of the theta function, restricted to $A(\widetilde X)$, give a “universal section” of the line bundles of that degree over $X$. This extends to the singular case a classical result of Riemann.
Read the full abstract
Differential Geometry
Contact Whirl Curves in Sasakian Lorentzian 3-Manifolds
We introduce and study contact whirl curves in three-dimensional Lorentzian contact manifolds, with emphasis on the Sasakian setting. This notion refines the concept of whirl curves by encoding the interaction between the adapted frame of a curve and the ambient contact structure through the Reeb vector field. For non-geodesic unit-speed contact whirl curves, we derive a differential equation governing the torsion in terms of the Frenet invariants and the contact data. In the Lorentzian Sasakian setting, this leads to rigidity phenomena of Lancret type. In particular, we prove that every non-geodesic Legendre Frenet curve is automatically a contact whirl curve, and consequently has constant torsion $\tau=1$. We also investigate the interaction between contact whirl curves and magnetic trajectories associated with the canonical contact magnetic field. We show that every non-geodesic curve which is simultaneously magnetic and contact whirl must be Legendre, and we obtain an explicit expression for its torsion in terms of the tensor $h=\frac12\mathcal L_\xi\Phi$. In the Sasakian case, this reduces to the universal law $\tau=1$. Finally, in the Lorentzian Heisenberg group endowed with its standard Sasakian structure, we derive a coordinate form of the whirl condition and use it to produce explicit examples, including a construction by quadratures of non-Legendre contact whirl curves and a horizontal helicoidal Legendre family.
Read the full abstract
Information Theory · Cross-listed from cond-mat.stat-mech
Field Theory of Data: Anomaly Detection via the Functional Renormalization Group. The 2D Ising Model as a Benchmark
We establish a correspondence between anomaly detection in high-noise regimes and the renormalization group flow of non-equilibrium field theories. We provide a physical grounding for this framework by proving that the detection of phase transitions in interacting non-equilibrium systems maps to the study of an effective equilibrium field theory near its Gaussian fixed point, which we identify with the universal Marchenko-Pastur distribution. Applying the Functional Renormalization Group to the two-dimensional Model A, we demonstrate that the noise-to-signal ratio acts as a physical temperature, where the signal emerges as ordered domains within a thermalized background of fluctuations. Using the exact Onsager solution as a benchmark, we show that this approach identifies critical thresholds with an error below 4%, significantly outperforming standard information-theoretic metrics such as the Kullback-Leibler divergence. Our results provide a universal strategy for resolving structures in complex datasets near criticality, bridging the gap between statistical mechanics and statistical inference.
Read the full abstract
Combinatorics
The alternating compositions of weighted differential operators yield the weights' Wronskian with which constant?
The alternated composition of $N=2p$ differential operators $w_j(x)\,\partial_x^p$ of strict order $p$ on the line $\mathbb{R}\ni x$ is again a differential operator of strict order $p$; its coefficient is the constant $\mathrm{const}(p)$, depending only on the arity $N$, times the Wronskian determinant of the originally taken coefficients $w_1,\dots,w_N$. The case $p=1$ of the Lie bracket for two vector fields fixes $\mathrm{const}(1)=1$, and $\mathrm{const}(2)=2$ is found easily by hand; $\mathrm{const}(3)=90$ can still be obtained symbolically. The problem is to determine $\mathrm{const}(p\geqslant4)$. We compute $\mathrm{const}(p)$ exactly for all $p\leqslant14$ – a 241-digit integer at $p=14$ – and record the resulting integer sequence as OEIS A392714. We prove that $v_p(\mathrm{const}(p))\geqslant p-1$ for every prime $p$, matching the exact equality observed numerically throughout our range, and conjecture that this equality holds in general. We show that $\log\mathrm{const}(p)$ grows like $\alpha p^2\log p$, with the leading coefficient close to $2$, nearly saturating the bound $p^2\log p\,(1+O(1/\log p))\leqslant\log\mathrm{const}(p)\leqslant2p^2\log p\,(1+O(1/\log p))$ obtained by O. Zaboronski (private communication). A naturally arising reduced constant is found to decay to zero super-exponentially rather than grow, a direct consequence of this near-saturation.
Read the full abstract
Differential Geometry · Cross-listed from cs.LG
Steerable Neural ODEs on Homogeneous Spaces
We introduce steerable neural ordinary differential equations on homogeneous spaces $M=G/H$. These models constitute a novel geometric extension of manifold neural ordinary differential equations (NODEs) that transport associated feature vectors transforming under the local symmetry group $H$. We interpret features as sections of associated vector bundles over $M$, and describe their evolution as parallel transport. This results in a coupled system of ODEs consisting of a flow equation on $M$ and a steering equation acting on features. We show that steerable NODEs are $G$-equivariant whenever the vector field generating the flow and the connection governing parallel transport are both $G$-invariant. Furthermore, we demonstrate how steerable NODEs incorporate existing NODE models and continuous normalizing flows on Lie groups. Our framework provides the geometric foundation for learning continuous-time equivariant dynamics of general vector-valued features on homogeneous spaces.
Read the full abstract
Geometric Topology
Thin surface subgroups of non-uniform arithmetic lattices in $\rm{SO}^+(n,1)$
We show that the fundamental groups of all non-compact, arithmetic, hyperbolic, $n$-manifolds for $n\geq 4$ contain thin surface subgroups. As a consequence of the proof of this theorem we also show that the fundamental groups of the doubles of cusped, arithmetic, hyperbolic $n$-manifolds embed as GFERF subgroups of $\rm{SO}^+(n+1,1)$.
Read the full abstract
Group Theory
Connections between the topology of the Morse boundary, the Morse local-to-global property and acylindrical hyperbolicity
We relate the topology of the Morse boundary of a group to geometric and algorithmic properties of the group. In particular, we show that a group has $\sigma$-compact Morse boundary if and only if it is Morse local-to-global. We also provide tools such as the geodesic Morse local-to-global property to show that groups are (not) Morse local-to-global. Our strategy generalizes tools from small cancellation theory, such as the intersection of relators, to arbitrary finitely generated groups. Further, we introduce a class of groups akin to graded small-cancellation groups and show that, for groups in this class, a geodesic is Morse if and only if its intersection with relators grows sublinearly in the length of the relators. We use this to construct the first example of a non-virtually cyclic Morse local-to-global group with an infinite-order Morse element that is not acylindrically hyperbolic.
Read the full abstract
Information Theory
Sensor Design for Accuracy-Bounded Estimation via Maximum-Entropy Likelihood Synthesis
Designing the sensing architecture for large-scale spatio-temporal systems is hard when accuracy requirements are specified but sensor models are uncertain or unavailable. Classical design treats sensor placement and estimation sequentially, requiring valid forward models for each sensing modality. This paper inverts the design flow: given an error budget, synthesize the measurement likelihood that enforces it while injecting minimal information beyond the dynamical prior. The likelihood is constructed by constrained optimization: among all posteriors satisfying a prescribed accuracy bound relative to a target, select the one minimizing Kullback-Leibler divergence from the prior. The solution is a maximum-entropy posterior in relative-entropy form, and the induced likelihood is the Radon-Nikodym derivative. The framework accommodates arbitrary discrepancies and is instantiated for Wasserstein distance, maximum mean discrepancy, $f$-divergences, moment constraints, and hybrid metrics. For each, we derive the discrete particle-level problem, analyze its convex or convex-relaxed structure, and present solvers with complexity scaling. A closed-form solution exists for the symmetric exponential-tilt case, and a distillation procedure converts nonparametric likelihood samples into parametric forms. A two-layer sensor design architecture embeds the synthesized likelihood in the recursive predict-update loop, connecting accuracy budgets to physical sensor placement, precision, and configuration. Numerical experiments comparing four metrics on unimodal and multimodal scenarios confirm the accuracy constraints are reliably enforced and reveal how metric choice determines the amount and spatial distribution of injected information.