Numerical Analysis
Solving Inverse Parametrized Problems via Finite Elements and Extreme Learning Networks
Erik Burman, Mats G. Larson, Karl Larsson et al. (+1 more)
We develop an interpolation-based modeling framework for parameter-dependent partial differential equations arising in control, inverse problems, and uncertainty quantification. The solution is discretized in the physical domain using finite element methods, while the dependence on a finite-dimensional parameter is approximated separately. We establish existence, uniqueness, and regularity of the parametric solution and derive rigorous error estimates that explicitly quantify the interplay between spatial discretization and parameter approximation. In low-dimensional parameter spaces, classical interpolation schemes yield algebraic convergence rates based on Sobolev regularity in the parameter variable. In higher-dimensional parameter spaces, we replace classical interpolation by extreme learning machine (ELM) surrogates and obtain error bounds under explicit approximation and stability assumptions. The proposed framework is applied to inverse problems in quantitative photoacoustic
tomography, where we derive potential and parameter reconstruction error estimates and demonstrate substantial computational savings compared to standard approaches, without sacrificing accuracy.
Read the full abstract ▾
All authors Erik Burman, Mats G. Larson, Karl Larsson, Jonatan Vallin
Geometric Topology
Flexible Surfaces in $\mathbb{C}P^2$ and $S^2\times S^2$
Joshua Lehman
A surface $\Sigma$ in a 4-
manifold $M$ is called flexible if any mapping class of the surface arises as the restriction of a diffeomorphism $(M,\Sigma) \to (M,\Sigma)$. We construct flexible surfaces in $\mathbb{C}P^2$ and $S^2 \times S^2$ within any prescribed non-characteristic homology class. Within characteristic homology classes there is a
spin structure obstructing flexibility and we construct so-called
spin-flexible representatives.
Read the full abstract ▾
Metric Geometry
A strengthening of the Blaschke-Santaló inequality for $o$-symmetric planar convex bodies
Károly J. Böröczky and Endre Makai Jr
We verify the inequality $$ \frac{|K|}{|E|}+\frac{|K^*|}{|E^*|}\leq 2 $$ for any $o$-symmetric convex body $K\subset\mathbb{R}^2$ where $E$ is either the John ellipse of maximal area contained in $K$ or the minimal area Löwner ellipse containing $K$. The analogous estimate may not hold if $K$ is a planar but the assumption of $o$-
symmetry is dropped, or if $K$ is $o$-symmetric convex body in $\mathbb{R}^n$ for $n\geq 3$. Our new inequality strengthens the Blaschke-Santaló inequality for $o$-symmetric convex bodies $K\subset\mathbb{R}^2$ with an error term of optimal order.
Read the full abstract ▾
Number Theory
Linear independence of theta series of positive-definite even unimodular lattices
Manuel K.-H. Müller
We show that the minimal $g$ for which the degree $g$ theta series of positive-definite even unimodular
lattices of rank $m\geq24$ are linearly independent is bounded between $\frac{m}{2}$ and $\frac{3m}{4}$.
Algebraic Geometry
Topological and arithmetic characteristics about products of projective lines with complex tori
Jia-Li Mo, Meirav Amram and Cheng Gong
In this paper, we study non-planar degeneracies with cylindrical configurations. They could be constructed by the product $\mathbb{CP}^1 \times T$ of the projective plane and a complex
torus with
embedding $(m,n)$. We prove that their fundamental groups of Galois covers have an abelian subgroup of rank $m(2n-1)$ respectively, and the irregularity of these surfaces are at least $2mn-1$. Furthermore, we also use Chern numbers to compute the index of such surfaces and classify them.
Read the full abstract ▾
Probability
Partial orders and monotonicity of logarithmic depth and height in preferential attachment trees
Christian Mönch
We study preferential attachment trees with general attachment rules. A natural intuition is that rewarding high-degree vertices more strongly should make the tree shallower. We test this intuition for the logarithmic growth constants of insertion depth and tree height. We show by an explicit counterexample that growth-ratio dominance alone does not guarantee either monotonicity. We identify additional tail-order conditions, arising from the branching-process representation of the tree, under which both expected comparisons hold.
Read the full abstract ▾
Mathematical Physics
Heterogeneous Cattaneo-Vernotte equation connection to the noisy voter model
K. Górska, A. Horzela, D. Jankov Maširević et al. (+3 more)
We consider a heterogeneous
diffusion equation and its corresponding generalization to the Cattaneo-Vernotte equation. It is derived by a combination of the continuity equation and the constitutive relation in various
stochastic interpretations of the heterogeneous
diffusion process. The heterogeneity in the system is introduced by considering a position-dependent
diffusion coefficient. Exact results for the probability density function and the mean squared displacement are provided. The limiting case of heterogeneous
diffusion is analyzed in detail, and the corresponding time-averaged mean-squared displacement is calculated. From the obtained results, an ergodicity breaking is observed.
Read the full abstract ▾
All authors K. Górska, A. Horzela, D. Jankov Maširević, T. Pietrzak, 1T. K. Pogány, T. Sandev
Functional Analysis
Cheeger's isoperimetric problem for Gaussian mixtures
Lukas Liehr
In any dimension $n$, we determine the Cheeger constant and the Cheeger sets of the Gaussian mixture $\mu(x) = p\gamma(x-a) + (1-p)\gamma(x-b)$, where $p \in [0,1]$, $a,b \in \mathbb{R}^n$, and $\gamma : \mathbb{R}^n \to (0,\infty)$ denotes a Gaussian. In particular, we characterize the Cheeger sets for $\mu$ in terms of specific half-spaces perpendicular to $a-b$, thereby confirming the conjectured solution to the Cheeger problem for Gaussian mixtures. Finally, we study the regime of parameters $p,a,b$ in which $\mu$ admits a unique Cheeger set.
Read the full abstract ▾
Probability
Hua-Pickrell diffusions and differential equations related with pseudo-Jacobi polynomials
Martin Auer and Michael Voit
Following Assiotis (2020), we study general $\beta$-Hua-Pickrell
diffusions of $N$ particles on $\mathbb R$ as solutions of the
stochastic differential equations (SDEs) $$dX_{j,t}=\sqrt{2(1+X_{j,t}^2)}\,dB_{j,t}+\beta\left[b-a X_{j,t}+\sum_{l=1,\ldots, N; \> l\neq j}\frac{X_{j,t}X_{l,t}+1}{X_{j,t}-X_{l,t}}\right]dt\,,\;\; (j=1,\ldots,N)$$ with $\beta\ge 1,\> a,b\in\mathbb R$. These processes form a subclass of the Pearson
diffusions which are defined as solutions of algebraic SDEs where the moments of the empirical distributions $\mu_t^N:=\frac{1}{N}\sum_{j=1}^N \delta_{X_{j,t}}$ can be computed inductively. This Pearson class also contains other well known
diffusions like Dyson Brownian motions, and multivariate Laguerre and Jacobi processes After the time normalization $t\mapsto t/\beta$, the SDEs above degenerate in the frozen case for $\beta=\infty$ into ordinary differential equations which are related to pseudo-Jacobi
polynomials. For $N\to\infty$ and under suitable initial conditions, the empirical distributions $\mu_t^N$ converge weakly almost surely for $t>0$ to some limit which is independent from $\beta\in[1,\infty]$. For $a=-N, b=0$, we describe the limit explicitly via free convolutions. Moreover, if $a=cN$ for some $c>0$, the solutions of our SDEs converge for $t\to\infty$ to stationary distributions, which are Hua-Pickrell (or Cauchy) measures. We thus obtain connections between known results for the empirical distributions of these ensembles and the zeros of the pseudo-Jacobi
polynomials. Furthermore, we derive a freezing central limit
theorem for $\beta\to\infty$ for the Hua-Pickrell ensembles which is related to these zeros.
Read the full abstract ▾
Number Theory
Completing the classification of torsion over sextic fields for elliptic curves defined over $\mathbb{Q}$
Nikola Adžaga and Tomislav Gužvić
We complete the classification of torsion subgroups $E(K)_{\text{tors}}$ occurring for an
elliptic curve $E/\mathbb{Q}$ over a sextic number field $K$. Previous work determined the complete set of possible groups up to one remaining case, namely $C_3 \oplus C_{18}$. We prove that this group does not occur.
Read the full abstract ▾
Combinatorics
Grid-free linear hypergraphs via Cayley-Bacharach
Cosmin Pohoata
We give a new construction showing that for every $r\ge 3$, there exists an $r$-uniform linear hypergraph on $n$ vertices with $\Theta_r(n^2)$ edges and no copy of the $r\times r$ grid. This complements the works of Füredi–Ruszinkó, Glock–Joos–Kim–Kühn–Lichev, Delcourt–Postle for $r \geq 4$, as well as the subsequent constructions of Gishboliner–Shapira and Solymosi for the case $r=3$.
Read the full abstract ▾
Mathematical Physics
Actions of Lie 2-algebras and comomentum maps
Philippe Bonneau, Véronique Chloup-Arnould, Angela Gammella et al. (+1 more)
In this paper we introduce the notion of a 2-action of a Lie 2-
algebra on an arbitrary
manifold M. Furthermore, in [Rog12], given a n-plectic
manifold (M, $\omega$), the authors consider a Lie Infinity-
algebra L$\infty$ (M, $\omega$), which is a higher analogue of the Poisson
algebra of observables associated to a
symplectic manifold. This Lie Infinity-
algebra reduces to a Lie 2-
algebra L^2 (M, $\omega$) when (M, $\omega$) is 2-plectic. Following ideas of N.L. Delgado [Del18], we introduce the Lie 2-
algebra D^2 (M, $\omega$), which generalises the Lie 2-
algebra L^2 (M, $\omega$) and its extension containing
Hamiltonian pairs. Given a two-plectic
manifold (M, $\omega$) and a Lie 2-
algebra g_1 $\oplus$ g_0 acting on M we define a comomentum map as a lift of the action, i.e., as a Lie 2-
algebra morphism from g_1 $\oplus$ g_0 to the extension of the Lie 2-
algebra D^2 (M, $\omega$). In an appendix, we discuss very explicitly numerous examples, classified according to their algebraic properties.
Read the full abstract ▾
All authors Philippe Bonneau, Véronique Chloup-Arnould, Angela Gammella, Tilmann Wurzbacher
Optimization and Control
Distributed Multi-Step Model Predictive Control for Consensus
Navid Noroozi
This paper studies consensus of discrete-time multi-
agent systems under time-varying directed communication, state and input constraints using a distributed multi-step model predictive control (MPC) framework. Consensus is recast as stabilization of the agreement set, and a
geometric viewpoint based on convex-hull invariance and strict interiority is adopted. Building on an existing
geometric necessary and sufficient condition for agreement, we show that enforcing terminal inclusion in local neighbor convex hulls guarantees hull invariance but does not, in general, imply the strict relative-interior property required for convergence. An explicit counterexample demonstrates that strictness cannot be deduced from feasibility and contraction constraints alone. To resolve this issue without shrinking feasible sets or altering primary performance objectives, a lexicographic tie-breaking mechanism is introduced. Among optimal (or near-optimal) MPC solutions, the proposed secondary criterion selects trajectories maximizing an interiority measure with respect to the neighbor hull. It is shown that whenever an interior feasible terminal state exists, this selection rule enforces the strictness condition required for asymptotic consensus. Explicit horizon conditions are derived for single- and double-integrator
agents with bounded inputs, ensuring feasibility and automatic existence of interior feasible terminal points. The resulting scheme provides a distributed and implementable route to consensus via finite-step set-Lyapunov contraction. Numerical simulations with distributed inter-process communication illustrate monotone diameter decay and report per-
agent computational complexity.
Read the full abstract ▾
Analysis of PDEs
Matching conditions for scattering solutions of scalar wave equations on extremal Reissner-Nordström spacetimes
Yannis Angelopoulos and Istvan Kadar
We study scattering solutions $\phi$ of the linear wave equation on extremal Reissner-Nordström spacetimes, satisfying the following properties: i) $\phi$ attains a prescribed radiation field $\psi_{\mathcal{I}}$ through future null infinity, which decays at an inverse
polynomial rate; ii) $\phi$ is regular in the exterior region up to and including the future event horizon, i.e. $\phi\in C^N$, where $N\gg1$ is independent of the decay rate of $\psi_{\mathcal{I}}$. We prove that such solutions exist for arbitrary $N$, and that they are not unique. The proof consists of: 1) finding an approximate solution $\phi_{\mathrm{app}}$ with fast decaying error; 2) the use of backwards energy estimates in order to correct $\phi_{\mathrm{app}}$ to an exact solution. Extremality is used only in the second step. The methods of the linear case described above are then used to show the same results for semilinear equations where the nonlinearity satisfies the null condition, as well as to geometries describing the
hyperbolic orbit of multiple extremal Reissner-Nordström black holes.
Read the full abstract ▾
Category Theory · Cross-listed from cs.DB
A Unified Mathematical Framework for Distributed Data Fabrics: Categorical Hypergraph Models
T. Shaska and I. Kotsireas
Current distributed data fabrics lack a rigorous mathematical foundation, often relying on ad-hoc architectures that struggle with consistency, lineage, and scale. We propose a mathematical framework for data fabrics, unifying heterogeneous data management in distributed systems through a hypergraph-based structure \( \mathcal{F} = (D, M, G, T, P, A) \).
Datasets, metadata, transformations, policies, and analytics are modeled over a distributed system \( \Sigma = (N, C) \), with multi-way relationships encoded in a hypergraph \( G = (V, E) \). A categorical approach, with
datasets as objects and transformations as morphisms, supports operations like data integration and
federated learning. The hypergraph is embedded into a modular tensor category, capturing relational
symmetries via braided monoidal structures, with
geometric analogies to Hurwitz spaces enriching the algebraic modeling. We prove the NP-hardness of critical tasks, such as schema matching and dynamic partitioning, and propose spectral methods and
symmetry-based
alignments for scalable solutions. The framework ensures consistency, completeness, and causality under CAP and CAL
theorems, leveraging sparse incidence
matrices and braiding actions for fault-tolerant operations.
Read the full abstract ▾
Symplectic Geometry
$L_\infty$-morphisms between twisted Courant $r$-Lie algebras and untwisted Courant $(r{+}1)$-Lie algebroids
Domenico Fiorenza and Antonio Michele Miti
In "Lie infinity
algebras and higher analogues of Dirac structures and Courant algebroids" [arXiv:1003.1004], Marco Zambon constructs an $L_\infty$-
algebra associated with any higher standard or twisted Courant algebroid (also known as a Vinogradov algebroid), and exhibits an explicit $L_\infty$-morphism from the Lie
algebra associated with a standard Lie algebroid twisted by a closed 2-form to the Lie-2
algebra of the standard Courant algebroid. He poses the question of whether analogous canonical $L_\infty$-morphisms exist in higher degrees – namely, for any standard higher Courant algebroid twisted by a closed $(r+1)$-form. We adfirmatively answer this question, presenting a general framework that naturally yields such canonical $L_\infty$-morphisms for arbitrary $r$, while at the same time clarifying the geometrical and homotopical structures underlying the construction. We also show how this framework accommodates the canonical morphism between Roger's observable $L_\infty$-
algebra of a pre-$r$-plectic
manifold and the higher Courant
algebra described by Zambon and one of the authors in "Observables on multisymplectic
manifolds and higher Courant algebroids" [arXiv:2209.05836].
Read the full abstract ▾
Probability · Cross-listed from stat.CO
Weak Poincaré inequalities for two-component Deterministic-scan Metropolis-within-Gibbs samplers
Mengxi Gao, Gareth O. Roberts and Andi Q. Wang
Using the framework of weak Poincaré inequalities, we analyze the convergence properties of deterministic-scan Metropolis-within-Gibbs samplers, an important class of Markov chain Monte Carlo
algorithms. Our analysis applies to nonreversible Markov chains and yields explicit (subgeometric) convergence bounds through novel comparison techniques based on Dirichlet forms. We show that the joint chain inherits the convergence behavior of the marginal chain and conversely. In addition, we establish several fundamental results for weak Poincaré inequalities for discrete-time Markov chains, such as a tensorization property for independent chains. We apply our theoretical results through applications to
algorithms for
Bayesian inference for a hierarchical regression model and a
diffusion model under discretely-observed data.
Read the full abstract ▾
Mathematical Physics · Cross-listed from quant-ph
Enhanced multiparameter quantum estimation in cavity magnomechanics via a coherent feedback loop
Adnan Naimy, Abdallah Slaoui, Abderrahim Lakhfif et al. (+1 more)
Multiparameter
quantum metrology plays a fundamental role in uncovering and exploiting the distinctive features of
quantum systems. In this work, we propose an effective and experimentally feasible scheme to significantly enhance the simultaneous
quantum estimation of the
photon magnon and
magnon mechanical coupling strengths in a hybrid cavity
magnon mechanical platform. Our approach relies on the assistance of a coherent feedback loop combined with the injection of a coherent driving field. We show that an appropriate tuning of the system and feedback parameters leads to a substantial reduction of the estimation errors associated with both coupling strengths. To quantify the metrological performance of the proposed scheme, we employ the
quantum Cramer Rao bound (QCRB) as a fundamental
benchmark for multiparameter estimation. We explicitly compute and compare the QCRBs derived from the symmetric logarithmic derivative (SLD) and the right logarithmic derivative (RLD) formalisms. Our results demonstrate that the RLD based QCRB is systematically lower than the SLD based bound, indicating superior estimation precision in the considered noncommutative estimation scenario. We further analyze the performance of heterodyne detection and show that, in suitable parameter regimes, the corresponding classical estimation precision closely approaches the ultimate
quantum limit predicted by our scheme. Finally, we discuss the experimental feasibility of the proposed setup within currently available cavity
magnon mechanical platforms. Owing to its general character, the framework developed here can be readily extended to the high precision estimation of other physical parameters in hybrid
quantum systems.
Read the full abstract ▾
All authors Adnan Naimy, Abdallah Slaoui, Abderrahim Lakhfif, Rachid Ahl Laamara
Analysis of PDEs
On the kinetic equation arising from the large-scale limit of the Cucker-Smale model
Ruicheng Cheng, Seung-Yeal Ha, Jaemoon Lee et al. (+1 more)
We propose a large-scale scaling viewpoint for deriving mesoscopic dynamics from interacting particle systems and apply it to the Cucker–Smale flocking model. In contrast with the classical mean-field regime leading to the Vlasov-type Cucker–Smale equation with spatially nonlocal (convolution)
alignment force, our scaling yields a kinetic equation whose
alignment field becomes local in space and nonlocal only in velocity. For the spatially homogeneous case, we obtain an explicit solution and derive quantitative flocking rates. For the spatially inhomogeneous equation we establish a local well-posedness in $W^{1,\infty}$ and in $C_b^{1,\alpha}$, highlighting the additional difficulties caused by the absence of a convolution structure. Moreover, for sufficiently small interaction strength we present a global well-posedness and a forward-in-time $L^1$ asymptotic completeness property. Finally, we investigate mono-kinetic solutions and exhibit finite-time blow-up scenarios.
Read the full abstract ▾
All authors Ruicheng Cheng, Seung-Yeal Ha, Jaemoon Lee, Zhenfu Wang
Mathematical Physics · Cross-listed from cs.CE
Identification of random material properties as stochastic inversion problem
Eliška Kočková and Anna Kučerová
Heterogeneity of many building materials complicates numerical modelling of structural behaviour. The material randomicity can be manifested by different values of material parameters of each material specimen. To capture inherent variability of heterogeneous materials, the model parameters describing the material properties are considered as random variables and their identification consists in solving a~
stochastic inversion problem. The
stochastic inversion is based on searching for probabilistic description of model parameters which provides the distribution of the model response corresponding to the distribution of the observed data. The paper presents two different formulations of the
stochastic inversion problem. The first formulation arises from the
Bayesian inference of uncertain statistical moments of a prescribed parameters' distribution while the main idea of the second one utilizes nonlinear transformation of random model parameters from distribution of the observed data.
Read the full abstract ▾
Optimization and Control
Joint Majorization-Minimization for Nonnegative CP and Tucker Decompositions under $\beta$-Divergences: Unfolding-Free Updates
Valentin Leplat
We study majorization-minimization methods for nonnegative tensor decompositions under the $\beta$-divergence family, focusing on nonnegative CP and Tucker models. Our aim is to avoid explicit mode unfoldings and large auxiliary
matrices by deriving separable surrogates whose multiplicative updates can be implemented using only tensor contractions (einsum-style operations). We present both classical block-MM updates in contraction-only form and a joint majorization strategy, inspired by joint MM for
matrix $\beta$-NMF, that reuses cached reference quantities across inexpensive inner updates. We prove tightness of the proposed majorizers, establish monotonic decrease of the objective, and show convergence of the sequence of objective values. For block-MM, we discuss how BSUM theory applies to the analysis of stationary accumulation points. For J-CoMM, we further establish, under a set of standard regularity assumptions and for one inner sweep per outer iteration, convergence of the iterates to a critical point through a KL-based analysis. Finally, experiments on synthetic tensors and the Uber spatiotemporal count tensor demonstrate substantial speedups over unfolding-based baselines and competitive runtime relative to a recent einsum-factorization framework.
Read the full abstract ▾
Optimization and Control · Cross-listed from cs.LG
Exposing Diversity Bias in Deep Generative Models: Statistical Origins and Correction of Diversity Error
Farzan Farnia, Mohammad Jalali and Azim Ospanov
Deep generative models have achieved great success in producing high-quality samples, making them a central tool across machine learning applications. Beyond sample quality, an important yet less systematically studied question is whether trained generative models faithfully capture the diversity of the underlying data distribution. In this work, we address this question by directly comparing the diversity of samples generated by state-of-the-art models with that of test samples drawn from the target data distribution, using recently proposed reference-free
entropy-based diversity scores, Vendi and RKE. Across multiple
benchmark datasets, we find that test data consistently attains substantially higher Vendi and RKE diversity scores than the generated samples, suggesting a systematic downward diversity bias in modern generative models. To understand the origin of this bias, we analyze the finite-sample behavior of
entropy-based diversity scores and show that their expected values increase with sample size, implying that diversity estimated from finite training sets could inherently underestimate the diversity of the true distribution. As a result, optimizing the generators to minimize divergence to empirical data distributions would induce a loss of diversity. Finally, we discuss potential diversity-aware regularization and guidance strategies based on Vendi and RKE as principled directions for mitigating this bias, and provide empirical evidence suggesting their potential to improve the results.
Read the full abstract ▾
Algebraic Topology
m-Contiguity Distance
Nilay Ekiz Yazici, Nursultan Kuanyshov and Ayse Borat
In this paper, we systematically develop the $m$-contiguity distance between simplicial maps as a discrete approximation framework for homotopical complexity in the category of simplicial complexes. We construct an increasing sequence of invariants that approximate the contiguity distance from below. We prove that $m$-contiguity distance is invariant under strong homotopy equivalence and that $m$-contiguity distance coincides with the usual contiguity distance provided that the dimension of the domain simplicial complex is $m$. The fundamental properties of $m$-contiguity distance are established, including its behaviour under barycentric subdivision, under compositions, and a categorical poduct inequality. As applications of this theory, we define the $m$-simplicial Lusternik-Schnirelmann category and the $m$-discrete topological complexity, proving that each arises naturally as a special case of $m$-contiguity distance. We also showed that $SD_1(\varphi,\psi)=SD(\varphi,\psi)$ under some conditions related to aspherical spaces.
Read the full abstract ▾
Geometric Topology
Uniform Perfectness and Centers in Sublinearly Morse Boundaries
Hyungryul Baik
The $\kappa$–Morse boundary was introduced for CAT(0) spaces by Qing and Rafi and extended to proper geodesic spaces by Qing, Rafi, and Tiozzo. Motivated by Han and Liu's work on uniformly perfect Morse boundaries, we ask when uniform perfectness of visual boundary data detects $\kappa$–center exhaustivity. Two locally finite trees show that fixed-basepoint uniform perfectness is insufficient and, when $\kappa$ is unbounded, a basepoint-independent absolute annular cutoff is not necessary. For locally finite trees, center exhaustivity is equivalent to fixed-basepoint uniform perfectness and $\kappa$–radial accessibility. Under explicit uniform visual-data hypotheses, analogous conditions imply center exhaustivity through a chosen boundary stratum in proper geodesic spaces; a normalized all-basepoint criterion is also obtained. Finally, we characterize the metric transforms $\phi$ for which one distortion function, depending only on $\phi$, works for every identity $(Z,d)\to(Z,\phi\circ d)$, and analyze the resulting metrics on rooted $q$–ary trees.
Read the full abstract ▾
Numerical Analysis · Cross-listed from cs.LG
Pseudo-differential-enhanced physics-informed neural networks
Andrew Gracyk
We present pseudo-differential enhanced physics-informed neural networks (PINNs), an extension of gradient enhancement but in
Fourier space. Gradient enhancement of PINNs dictates that the PDE residual is taken to a higher differential order than prescribed by the PDE, added to the objective as an augmented term in order to improve training and overall learning fidelity. We propose the same procedure after application via
Fourier transforms, since differentiating in
Fourier space is multiplication with the
Fourier wavenumber under suitable decay. Our methods are fast and efficient. Our methods oftentimes achieve superior PINN versus numerical error in fewer training iterations, potentially pair well with few/fixed samples in collocation, and can achieve breakthrough rather than gradual effects on the error. Moreover, our methods are suitable for fractional derivatives. We establish that our methods, due to the dynamical effects, improve spectral
eigenvalue decay of the neural tangent kernel (NTK), and so our methods contribute towards the learning of high frequencies in early training, mitigating the effects of frequency bias up to the
polynomial order and possibly greater with smooth activations. In particular, our primary contribution can be characterized as gradient enhancement affects spectral bias, and we specialize to
Fourier space, although our theory does not prove the ambient case exactly due to a result with the Plancherel
theorem on the descent paths. Our methods accommodate advanced techniques in PINNs, such as
Fourier feature
embeddings. A pitfall of discrete
Fourier transforms via the Fast
Fourier Transform (FFT) is mesh subjugation, and so we demonstrate compatibility of our methods for greater mesh flexibility and invariance on alternative Euclidean and non-Euclidean domains via Monte Carlo methods and otherwise, although possessing trade-offs of their own.
Read the full abstract ▾
Analysis of PDEs
Morrey estimates for the gradient in non-linear variational transmission problems
Luca Esposito and Lorenzo Lamberti
We study a class of variational transmission problems driven by nonlinear energies with discontinuous coefficients across a prescribed interface. The model setting consists of integral functionals of the form \[ \mathcal{F}(u;E)=\int_{\Omega}\sigma_E(x)\,F(\nabla u)\,dx, \] where the coefficient $\sigma_E$ takes two constant values on complementary regions separated by a $C^1$ hypersurface, and the integrand $F$ satisfies standard $p$-growth and monotonicity conditions with $p>2$. In this nonlinear variational framework, we establish local Morrey-space regularity for the gradient of local minimizers, proving that $\nabla u\in L^{2,\lambda}_{\mathrm{loc}}(\Omega)$ for every $0\leq\lambda<n$, provided $2<p<\frac{2n}{n-2}$. The proof is based on quantitative decay estimates for the energy near the interface, first obtained in a flat configuration and then extended to the general case by a suitable approximation argument.
Read the full abstract ▾
Algebraic Geometry
Computing $A$-resultants via direct images
Friedemann Groh and Matthias Zach
We improve a previously known theoretic method to compute A-resultants for suitable monomial support sets due to Weyman to the extent that it becomes computationally feasible and effective. This is achieved by introducing a new
algorithm for the computation of direct images of complexes of coherent sheaves on toric varieties. The procedure does not rely on Gröbner basis computations at any stage.
Read the full abstract ▾
Differential Geometry · Cross-listed from cs.LG
An Embarrassingly Simple Way to Optimize Orthogonal Matrices at Scale
Adrián Javaloy and Antonio Vergari
Orthogonality constraints are ubiquitous in robust and probabilistic machine learning. Unfortunately, current optimizers are computationally expensive and do not scale to problems with hundreds or thousands of constraints. One notable exception is the Landing
algorithm (Ablin et al., 2024) which, however comes at the expense of temporarily relaxing orthogonality. In this work, we revisit and improve on the ideas behind Landing, enabling the inclusion of modern adaptive optimizers while ensuring that orthogonal constraints are effectively met. Remarkably, these improvements come at little to no cost, and reduce the number of required hyperparemeters. Our
algorithm POGO is fast and GPU-friendly, consisting of only 5
matrix products, and in practice maintains orthogonality at all times. On several challenging
benchmarks, POGO greatly outperforms recent optimizers and shows it can optimize problems with thousands of orthogonal
matrices in minutes while alternatives would take hours. As such, POGO sets a milestone to finally exploit orthogonality constraints in ML at scale. A PyTorch implementation of POGO is publicly available at https://github.com/adrianjav/pogo.
Read the full abstract ▾
Optimization and Control
Temporally Flexible Transport Scheduling on Networks with Departure-Arrival Constriction and Nodal Capacity Limits
Anqi Dong, Karl H. Johansson and Johan Karlsson
We investigate the optimal transport (OT) problem over networks, wherein supply and demand are conceptualized as temporal marginals governing departure rates of particles from source nodes and arrival rates at sink nodes. This setting extends the classical OT framework, where all mass is conventionally assumed to depart at $t = 0$ and arrive at $t = t_f$. Our generalization accommodates departures and arrivals at specified times, referred as departure–arrival(DA) constraints. In particular, we impose nodal-temporal flux constraints at source and sink nodes, characterizing two distinct scenarios: (i) Independent DA constraints, where departure and arrival rates are prescribed independently, and (ii) Coupled DA constraints, where each particle's transportation time span is explicitly specified. We establish that OT with independent DA constraints admits a multi-marginal optimal transport formulation, while the coupled DA case aligns with the unequal-dimensional OT framework. For line graphs, we analyze the existence and uniqueness of the solution path. For general graphs, we use a constructive path-based reduction and optimize over a prescribed set of paths. From a computational perspective, we consider entropic regularization of the original problem to efficiently provide solutions based on multi-marginal Sinkhorn method, making use of the graphical structure of the cost to further improve scalability. Our numerical simulation further illustrates the linear convergence rate in terms of marginal violation.
Read the full abstract ▾
Differential Geometry
Asymptotic behavior at infinity of Weingarten surfaces
Aires E. M. Barbieri, José A. Gálvez, Yuanyuan Lian et al. (+1 more)
We derive the asymptotic expansion at infinity for embedded ends of uniformly
elliptic Weingarten surfaces with finite total
curvature in $\mathbb{R}^3$, and we establish a maximum principle at infinity. Furthermore, we solve the Dirichlet problem for the uniformly
elliptic Weingarten equation in dimension two on strictly convex bounded domains.
Read the full abstract ▾
All authors Aires E. M. Barbieri, José A. Gálvez, Yuanyuan Lian, Kai Zhang