Numerical Analysis · Cross-listed from physics.geo-ph
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
Monday 16 March
Marine biogeochemical models are widely used to study nutrient dynamics, water quality, and climate-related processes in coastal and estuarine systems. However, developing models that reliably represent specific environments remains computationally demanding, which makes their application to complex systems such as river plumes and estuarine environments challenging. In addition, these models contain several parameters that must be calibrated for the region of interest, a process that is often performed empirically using limited observational data. This thesis advances the development and calibration of marine biogeochemical models in the Brazilian context through three main contributions. First, we develop a conceptual model describing nutrient-phytoplankton dynamics in the Paranagua Estuarine Complex (PEC) in southern Brazil. The model is intentionally simple and computationally inexpensive, allowing simulations to be performed on standard personal computers. Second, we propose a systematic calibration framework based on tracer datasets and derivative-free optimization techniques. Finally, we demonstrate the practical application of this approach by calibrating the PEC model using in situ observations. Results show that, despite its simplicity, the model can reproduce observed nutrient dynamics when properly calibrated. The proposed framework is general and can be extended to multi-parameter calibration, seasonal parameter variation, and the coupling of biogeochemical models with higher-fidelity hydrodynamic models.
Algebraic Geometry
On some invariants of hypersurface singularities
Given a hypersurface defined by $f$ in a smooth complex algebraic variety $X$, and a point $P$ on this hypersurface, we consider the invariant $\beta_P(f)$ given by the log canonical threshold at $P$ of ${\mathfrak m}_P\cdot J_f$, where ${\mathfrak m}_P$ is the ideal defining $P$ and $J_f$ is the Jacobian ideal of $f$. We show that this invariant satisfies most of the formal properties of the log canonical threshold of $f$ and give some examples. Dano Kim asked whether this invariant always gives an upper bound for the minimal exponent of $f$ at $P$. Motivated by this, we raise another question about minimal exponents, give a positive answer to a weaker version, and discuss some examples.
Read the full abstract
Optimization and Control · Cross-listed from cs.LG
Unbiased and Biased Variance-Reduced Forward-Reflected-Backward Splitting Methods for Stochastic Composite Inclusions
This paper develops new variance-reduction techniques for the forward-reflected-backward splitting (FRBS) method to solve a class of possibly nonmonotone stochastic composite inclusions. Unlike unbiased estimators such as mini-batching, developing stochastic biased variants faces a fundamental technical challenge and has not been utilized before for inclusions and fixed-point problems. We fill this gap by designing a new framework that can handle both unbiased and biased estimators. Our main idea is to construct stochastic variance-reduced estimators for the forward-reflected direction and use them to perform iterate updates. First, we propose a class of unbiased variance-reduced estimators and show that increasing mini-batch SGD, loopless-SVRG, and SAGA estimators fall within this class. For these unbiased estimators, we establish a $\mathcal{O}(1/k)$ best-iterate convergence rate for the expected squared residual norm, together with almost-sure convergence of the iterate sequence to a solution. Consequently, we prove that the best oracle complexities for the $n$-finite-sum and expectation settings are $\mathcal{O}(n^{2/3}\epsilon^{-2})$ and $\mathcal{O}(\epsilon^{-10/3})$, respectively, when employing loopless-SVRG or SAGA, where $\epsilon$ is a desired accuracy. Second, we introduce a new class of biased variance-reduced estimators for the forward-reflected direction, which includes SARAH, Hybrid SGD, and Hybrid SVRG as special instances. While the convergence rates remain valid for these biased estimators, the resulting oracle complexities are $\mathcal{O}(n^{3/4}\epsilon^{-2})$ and $\mathcal{O}(\epsilon^{-5})$ for the $n$-finite-sum and expectation settings, respectively. Finally, we conduct two numerical experiments on AUC optimization for imbalanced classification and policy evaluation in reinforcement learning.
Read the full abstract
General Topology
Are scales Fréchet?
We continue the study of Dow spaces of a $\mathfrak{b}$-scale, originally introduced by Alan Dow in "$\pi$-Weight and the Fréchet-Urysohn property" (Topology and its Applications, Vol. 174, pp. 56-61). We prove that it is consistent that all such spaces are Fréchet, but it is also consistent that none of them is. We use these spaces to exhibit (consistently) a $\triangle_{2}^{1}$ ideal that does not satisfy the Category Dichotomy. Finally, we prove that the Category Dichotomy holds for all co-analytic ideals.
Read the full abstract
Complex Variables
A curvature estimate for holomophic maps on open Riemann surfaces
We apply the technique of jet differentials to establish a Gauss curvature estimate for an open Riemann surface $M$, equipped with a conformal metric induced from a nonconstant holomorphic map that is highly ramified over a generic hypersurface of sufficiently high degree.
Number Theory
Diversity, equity, and inclusion for problems in additive number theory
This is a survey of the diversity of problems in additive number theory. Equity requires the consideration of less currently popular problems, and suggests their inclusion in the additive canon. Of particular interest are problems about the sizes of sumsets of finite sets of integers and problems about the arithmetical structure of intersections of sumsets.
Read the full abstract
Analysis of PDEs
Boundary symmetry breaking via logistic damping in a chemotaxis-growth system
We establish global stability for a chemotaxis-growth model with logarithmic sensitivity under dynamic Dirichlet boundary conditions on a 1D domain. We analyze both parabolic-parabolic and parabolic-hyperbolic systems. The key challenge is handling time-dependent boundary data for the unknown functions. We overcome this by introducing dynamic reference profiles which suitably interpolate boundary values. Using an expanded entropy functional measuring deviation from these profiles, we prove energy estimates the uniform boundedness of solutions and global asymptotic stability of perturbations.
Read the full abstract
Combinatorics
A Recognizable Substitution for the Anti-Penrose Tiling
We present a substitution on six marked rhombs, called Seabed, with inflation factor $\varphi^3$. We prove that its tiling space is mutually locally derivable with the unmarked Anti-Penrose pentagrid hull. Eleven vertex patches of depth at most four recognize the parent vertices: their acceptance domains coincide with the five internally contracted windows. Matching the fine tiles around each parent rhomb to a replacement patch recovers their markings uniquely. The inverse local maps have a uniform finite radius and extend to all singular limit tilings in the hull. The construction also gives a recognizable parent hierarchy.
Read the full abstract
Number Theory
Lifts of unramified twists and local-global principles
We prove that two-step nilpotent $p$-extensions of rational global function fields of characteristic $p$ satisfy a quantitative local-global principle when they are counted according to their largest upper ramification break ("last jump"). We had previously shown this only for $p\neq2$. Compared to our previous proof, this proof is also more self-contained, and may apply to heights other than the last jump. As an application, we describe the distribution of last jumps of $D_4$-extensions of rational global function fields of characteristic $2$. We also exhibit a counterexample to the analogous local-global principle when counting by discriminants.
Read the full abstract
Combinatorics
Drazin Inverses and Walk Structure of Oriented Dutch Windmill Graphs
We investigate the Drazin invertibility of adjacency matrices associated with a class of oriented graphs known as oriented Dutch windmill graphs. By analyzing walks of prescribed lengths and exploiting the structure of the minimal polynomial, we obtain explicit expressions for the Drazin inverse and determine its index. The approach combines combinatorial enumeration with algebraic matrix analysis, offering a constructive characterization that generalizes known results for paths, cycles, and bipartite graphs. Beyond its intrinsic theoretical value, the framework provides insight into discrete models governed by cyclic feedback and may serve as a basis for symbolic computation of generalized inverses in structured networks.
Read the full abstract
Functional Analysis
$q$-Numerical Ranges and Spectral Sets
We study spectral constants for convex domains $\Omega$ containing the spectrum of a matrix. We extend the Crouzeix–Palencia framework by obtaining bounds depending on a parameter $\gamma$ and relating these bounds to geometric properties of $\Omega$ and the numerical range $W(A)$. We generalise the proof that the numerical range is a $1+\sqrt{2}$-spectral set to scaled $q$-numerical ranges. We also formulate a generalisation of Crouzeix's Conjecture in the context of $q$-numerical ranges and give an application of this result to GMRES convergence bounds.
Read the full abstract
Optimization and Control
Notes on the primal-dual algorithm for convex optimization applied to X-ray tomographic image reconstruction
The purpose of these notes is to provide background on understanding the primal-dual algorithm of Chambolle and Pock [1] for imaging scientists. The presentation focuses on providing intuition and an algorithmic system that is amenable to pre-conditioning. The document aims to be self-contained, providing background on the essential facts of non-smooth convex analysis.[2]
Read the full abstract
Algebraic Geometry
On the separation Łojasiewicz exponents of real analytic sets in the real plane
The main aim of the paper is to give a formula for computing the separation Łojasiewicz exponents for two real analytic set germs via the Newton–Puiseux expansions of their defining functions. Moreover, we present an effective exponent for the case of two real algebraic sets in terms of their degrees.
Read the full abstract
Group Theory
Annexes in affine Coxeter complexes
We introduce the annex of an element $x$ in a Coxeter group as the set of elements $y$ such that $x \nleq y$ with respect to Bruhat order. This notion provides a complementary perspective to the study of Bruhat intervals and their interpretation via folded galleries. We establish general properties of annexes and show that in affine Coxeter groups the annex of any fixed element is finite. In rank-two affine Coxeter complexes, we further describe the geometric structure of annex boundaries using descent sets and configurations of parallel reflections. These results offer a new geometric viewpoint on the structure of the Bruhat order.
Read the full abstract
Optimization and Control · Cross-listed from q-fin.PM
Some general results on risk budgeting portfolios
Given a reference risk measure, the risk budgeting is the portfolio where each asset contributes a predetermined amount to the total risk. We propose a novel approach, alternative to the ones proposed in the literature, for the calculation of the risk budgeting portfolio. This different perspective on the problem has several interesting consequences. For the calculation of the portfolio, we define a Cauchy sequence within the simplex of R^n, whose limit corresponds to the risk budgeting portfolio. This construction allows for the straightforward implementation of an efficient algorithm, avoiding the need to solve auxiliary, equivalent optimization problems, which may be computationally challenging and hard to interpret in the decision theory context. We compare our algorithm with the standard optimization-based methods proposed in the literature. From a theoretical point of view, starting from the Cauchy sequence, we define a function for which the risk budgeting portfolio is a fixed point. Therefore, sufficient conditions for the existence and uniqueness of the fixed point can be used. The methodology is developed for general risk measures and implemented in detail in the case of standard deviation.
Read the full abstract
Optimization and Control
A Technical Note on the Implementation and Use of PDCS
This technical note documents the implementation and use of the Primal-Dual Conic Programming Solver (PDCS), a first-order solver for large-scale conic optimization problems introduced by Lin et al. (arXiv:2505.00311). It describes the algorithmic and implementation details underlying PDCS, including the restarted primal-dual hybrid gradient method framework, adaptive step-size selection, adaptive reflected Halpern iterations, adaptive restarts, and diagonal preconditioning. It also provides practical instructions for using PDCS, including its interfaces with JuMP and CVXPY, solver options, and illustrative code examples. PDCS is available at https://github.com/ZikaiXiong/PDCS under the Apache License 2.0.
Read the full abstract
Algebraic Geometry
Logarithmic spectral correspondence for $V$–twisted Higgs bundles on punctured curves
Let $X$ be a smooth projective complex curve, $P\subset X$ a reduced effective divisor, and $X^{0}=X\setminus P$. We study logarithmic $V$-twisted Higgs bundles arising from a logarithmic Hecke compactification of a rank-two bundle on $X^{0}$. We show that a pair of induced logarithmic line-twisted fields lifts uniquely exactly under explicit local Hecke conditions, and that the lift is integrable precisely when the fields commute. Fixing the compactified spectral curve $Y$, we classify such Higgs bundles by pairs $(F,\,\vartheta)$, where $F$ is a rank-one torsion-free sheaf on $Y$ and $\vartheta$ satisfies a marked spectral condition on a finite subscheme $Z\subset Y$. This gives a logarithmic extension of the compact rank-two spectral correspondence of~\cite{ABK} to the punctured case. On the line-bundle locus, the moduli stack is canonically equivalent to $\mathrm{Pic}^{d}(Y)\times A_Z$.
Read the full abstract
Optimization and Control
A practical randomized trust-region method to escape saddle points in high dimension
Without randomization, escaping the saddle points of $f \colon \mathbb{R}^d \to \mathbb{R}$ requires at least $\Omega(d)$ pieces of information about $f$ (values, gradients, Hessian-vector products). With randomization, this can be reduced to a polylogarithmic dependence in $d$. The prototypical algorithm to that effect is perturbed gradient descent (PGD): through sustained jitter, it reliably escapes strict saddle points. However, it also never settles: there is no convergence. What is more, PGD requires precise tuning based on Lipschitz constants and a preset target accuracy. To improve on this, we modify the time-tested trust-region method with truncated conjugate gradients (TR-tCG). Specifically, we randomize the initialization of tCG (the subproblem solver), and we prove that tCG automatically amplifies the randomization near saddles (to escape) and absorbs it near local minimizers (to converge). Saddle escape happens over several iterations. Accordingly, our analysis is multi-step, with several novelties. The proposed algorithm is practical: it essentially tracks the good behavior of TR-tCG, with three minute modifications and a single new hyperparameter (the noise scale $\sigma$). We provide an implementation and numerical experiments.
Read the full abstract
Optimization and Control
Minimal enclosing balls via geodesics
Algorithms for minimal enclosing ball problems are often geometric in nature. To highlight the metric ingredients underlying their efficiency, we focus here on a particularly simple geodesic-based method. A recent subgradient-based study proved a complexity result for this method in the broad setting of geodesic spaces of nonpositive curvature. We present a simpler, intuitive and self-contained complexity analysis in that setting, which also improves the convergence rate. We furthermore derive the first complexity result for the algorithm on geodesic spaces with curvature bounded above.
Read the full abstract
Probability
Explicit Solution of Infinite-Horizon Linear Backward Stochastic Volterra Integral Equations
We study linear backward stochastic Volterra integral equations (BSVIEs) on the infinite time horizon. By introducing weighted function spaces with exponential decay, we establish existence and uniqueness of adapted M-solutions. We construct an infinite-horizon resolvent kernel and derive explicit formulas for the solution components (Y,Z,K) using a Girsanov transformation and Hida Malliavin calculus. The results extend the finite-horizon theory of Hu and Oksendal to the infinite horizon framework.
Read the full abstract
Probability
Mean Field Games with Reflected Dynamics: Penalization and Relaxed Control Approach
In this paper, we investigate a class of Mean Field Games (MFGs) in which the state dynamics are governed by multidimensional reflected stochastic differential equations (SDEs). We establish the existence of an equilibrium and show that it can be approximated by the equilibrium of MFGs with non-reflected SDE. This approximation is constructed via a penalization method combined with the relaxed control approach introduced in [21]. Under a uniform ellipticity condition, and by applying the penalization method together with the mimicking theorem, we prove the existence of a Markovian MFG. Furthermore, under an additional convexity assumption, we demonstrate the existence of a strict-Markovian MFG. In the general case, we prove that relaxed MFG solutions with reflected dynamics can be approximated by strict controls whose dynamics are governed by penalized SDEs.
Read the full abstract
Information Theory
DMD Prediction of MIMO Channel Using Tucker Decomposition
Accurate channel state information (CSI) prediction is crucial for next-generation multiple-input multiple-output (MIMO) communication systems. Classical prediction methods often become inefficient for high-dimensional and rapidly time-varying channels. To improve prediction efficiency, it is essential to exploit the inherent low-rank tensor structure of the MIMO channel. Motivated by this observation, we propose a dynamic mode decomposition (DMD)-based prediction framework operating on the low-dimensional core tensors obtained via a Tucker decomposition. The proposed method predicts reduced-order channel cores, significantly lowering computational complexity. Simulation results demonstrate that the proposed approach preserves the dominant channel dynamics and achieves high prediction accuracy.
Read the full abstract
Dynamical Systems
A universal model for the bifurcations of asymptotic values
We study a universal model for the bifurcation of asymptotic values, based on the notion of tangent-like maps and families thereof. In analogy with polynomials and polynomial-like maps, we prove that every tangent-like map can be associated to a member of the model family. In particular, the filled Julia sets of tangent-like maps are quasiconformally homeomorphic to the ones for the model family. Likewise, the bifurcation locus of a family of tangent-like maps is homeomorphic to the boundary of the Tandelbrot set $\mathcal T$, the bifurcation locus of the model. Moreover $\mathcal T$ is universal, in the sense that homeomorphic copies of it can be found densely in the activity loci of asymptotic values of any appropriate natural family of finite type meromorphic maps. When applied to the iterates of a transcendental meromorphic map, these results make it possible to build a transcendental renormalization theory, a question that has been open for the last 25 years.
Read the full abstract
Group Theory
Variations of Demushkin Groups that are not Absolute Galois Groups
We construct two families of examples of pro-p groups, with rather elementary presentations, that do not complete into 1-cyclotomic oriented pro-p groups. These provide brand new examples of pro-p groups that do not occur as maximal pro-p Galois groups of fields containing a root of unity of order p - and thus, as absolute Galois groups. Moreover, we show that these pro-p groups may not be ruled out as maximal pro-p Galois groups employing other cohomological properties that are known to hold for all maximal pro-p Galois groups, such as the triple Massey vanishing property, or the quadraticity of Fp-cohomology.
Read the full abstract
Combinatorics · Cross-listed from cs.DS
The Compilability Thresholds of 2-CNF to OBDD
We prove the existence of two thresholds regarding the compilability of random 2-CNF formulas to OBDDs. The formulas are drawn from $\mathcal{F}_2(n,\delta n)$, the uniform distribution over all 2-CNFs with $\delta n$ clauses and $n$ variables, with $\delta \geq 0$ a constant. We show that, with high probability, the random 2-CNF admits OBDDs of size polynomial in $n$ if $0 \leq \delta < 1/2$ or if $\delta > 1$. On the other hand, for $1/2 < \delta < 1$, with high probability, the random $2$-CNF admits only OBDDs of size exponential in $n$. It is no coincidence that the two “compilability thresholds” are $\delta = 1/2$ and $\delta = 1$. Both are known thresholds for other CNF properties, namely, $\delta = 1$ is the satisfiability threshold for 2-CNF while $\delta = 1/2$ is the treewidth threshold, i.e., the point where the treewidth of the primal graph jumps from constant to linear in $n$ with high probability.
Read the full abstract
Information Theory
Enabling mmWave Communications with VCSEL-Based Light-Emitting Reconfigurable Intelligent Surfaces
This paper proposes a light-emitting reconfigurable intelligent surface (LeRIS) architecture that integrates vertical cavity surface-emitting lasers (VCSELs) to jointly support user localization and mmWave communication. By leveraging the directional Gaussian beams and dual-mode diversity of VCSELs, we derive a closed-form method for estimating user position and orientation using only three VCSEL sources. These estimates are then used to configure LeRIS panels for directional mmWave beamforming, enabling optimized wave propagation in programmable wireless environments. Simulation results demonstrate that the proposed system achieves millimeter-level localization accuracy and maintains high spectral efficiency. These findings establish VCSEL-integrated LeRIS as a scalable and multifunctional solution for future 6G programmable wireless environments.
Read the full abstract
Combinatorics
Quantized rational chip-firing
This article introduces a quantized chip-firing model with close connections to the theory of rational lattice paths and rational parking functions. Given a graph with a sink and positive integers a,b,c with gcd(a,b)=1, a set S of vertices fires by the following rule. Each vertex in S provisionally sends c chips to the sink and a/b chips to each non-sink neighbor outside of S. The novel feature is that the total number of chips leaving from or arriving at any vertex gets rounded down to the nearest integer before being finalized. We define the notions of chip configurations being superstable, k-stable, or k-skeletal in this model. When c=1 and the graph is complete, superstable configurations correspond to rational parking functions. There is a bijection between superstable configurations and k-skeletal configurations for each k. We establish these results by building a combinatorial theory of k-skeletal rational lattice paths (both unlabeled and labeled) and translating that theory to chip configurations. There is a group structure on the set of chip configurations modulo firing and borrowing moves. We show that this group is isomorphic to the product of b-1 copies of the integers modulo a; and, for each k, each coset of chip configurations in this group contains a unique k-skeletal representative.
Read the full abstract
Number Theory
$p$-adic $L$-functions for Hecke characters of totally imaginary fields
We construct $p$-adic $L$-functions interpolating critical $L$-values of algebraic Hecke characters for arbitrary unramified primes $p$ and any totally imaginary field. For non-ordinary primes, the only previously known case was that of imaginary quadratic extensions of $\mathbb{Q}$. One of the main ingredients is a new $p$-adic Fourier theory relating generic fibers of $p$-divisible groups to a general class of character varieties. Combining this with equivariant cohomology classes constructed in a previous paper allows us to construct the $p$-adic $L$-function.
Read the full abstract
Algebraic Geometry
Terminal Fano 4-fold Hypersurfaces via Deep Reinforcement Learning
We construct $46{,}992$ new families of Fano $4$-fold hypersurfaces $X_d\subset\mathbb(a_1,\ldots,a_6)$ of Fano index $1$ with $d>200$, whose general members have terminal singularities. Unlike the $11{,}617$ quasismooth families classified by Brown and Kasprzyk, these are nonquasismooth, for which determining terminality can require the analysis of hyperquotient singularities. An exhaustive search has classified all nonquasismooth families with $d\leq200$, but extending this classification to higher degrees becomes computationally prohibitive. We therefore exploit the sparsity and spatial clustering of terminal examples using two best-first searches over $\mathbb{Z}^6$: one with a fixed heuristic, and the other with a neural network trained online by deep reinforcement learning. Together, the searches reach degrees up to $3{,}234$. The neural network additionally discovers an average of $79$ hard-to-reach families per run. For the most isolated examples, our conservative distance-based model gives an estimated lower bound of nearly $40\%$ of the fixed search's total step budget to discover a single such family. An ablation study demonstrates that this advantage arises from the neural network rather than stochastic exploration alone. The complete list of families is publicly available.
Read the full abstract
Numerical Analysis · Cross-listed from cs.LG
Physics-informed fine-tuning of foundation models for partial differential equations
Foundation models for partial differential equations (PDEs) have emerged as powerful surrogates pre-trained on diverse physical systems, but adapting them to new downstream tasks remains challenging due to limited task-specific data and distribution shifts. While fine-tuning has proven transformative in natural language processing, best practices for adapting PDE foundation models remain underexplored. Although physics-informed training has successfully trained accurate solvers across a wide range of PDE problems, its potential for fine-tuning data-based foundation models has not been systematically studied. In this work, we introduce a physics-informed fine-tuning framework that adapts pre-trained PDE foundation models by incorporating physical constraints (PDE residuals and boundary conditions) directly into the fine-tuning objective. This enables effective adaptation in data-scarce regimes while promoting physical consistency. We evaluate our method on a downstream task composed of an unseen PDE class and compare it with data-driven finetuning counterparts. Our results demonstrate that physics-informed fine-tuning achieves competitive accuracy without requiring PDE solutions for training. Furthermore, a hybrid fine-tuning strategy yields superior generalization to out-of-distribution scenarios when only minimal training data is available. These findings establish physics-informed fine-tuning as a scalable and data-efficient paradigm, providing a physically interpretable pathway for adapting foundation models in scientific machine learning.