Papers

We develop regularisation theory for dynamic inverse problems, solved using online methods with an infinite time horizon. Using concepts of subregularity to treat nonsmooth regularisers, we prove that time-averaged reconstruction errors converge to zero as noise, algorithmic errors, and regularisation vanish as the horizon grows. We illustrate the theory numerically with a dynamic electrical impedance tomography example.

Let $\mathcal F$ denote the fundamental domain for $\text{SL}_2(\mathbb{Z})$ on the upper half plane $\mathcal H$. William Duke showed that as fundamental discriminants $D \to -\infty$, the sets $\mathrm{CM}_{D}$ (CM points of discriminant $D$) are equidistributed in $\mathcal F$. In this paper, we investigate the behavior of CM points on the boundary of $\mathcal F$. We prove that such CM points are equidistributed on the boundary, and also give a complete characterization of when every $\mathrm{CM}_D$ point lies on the boundary. Along the way, we also (conditionally) give a complete classification of negative discriminants with class group of small exponent.

We prove a nonterminating well-poised basic $q$-Taylor expansion with complementary remainders for a two-basis infinite-product kernel implicitly proposed by the second author in \cite[Sec.~5]{Schlosser2008}. The well-poised parameter $c$ gives the rational $p=0$ basis, while the elliptic nome $p$ is a separate deformation; the infinite expansions treated here are specific to the basic case. We compute the two Taylor coefficient families and show that each one-family Taylor remainder tends to the complementary basis contribution. The proof uses the well-poised Cooper formula, Jackson's terminating ${}_8\phi_7$ summation, Rogers' ${}_6\phi_5$ summation, and theta interpolation, but not Bailey's nonterminating ${}_8\phi_7$ summation, which is recovered as a consequence. We also record two quadratic one-family examples and discuss a multi-kernel outlook.

Motivated by a recent random pipe dream model, we study a family of probability distributions on \(S_n\) arising from Bott–Samelson varieties over finite fields. More precisely, for a word \(R\), we consider the Bott–Samelson map \(\pi_R:\mathrm{BS}^R\to \mathcal{F}\ell_n\) and define a distribution \(\mathbb{P}_{R,q}\) by counting the \(\mathbb{F}_q\)-points in the inverse images of Schubert cells. For a suitable choice of parameters \(p_1=q/(1+q)\) and \(p_2=1/q\), this construction recovers a special case of the random pipe dream distribution. The main problem considered in this note is to determine which combinatorial properties of a reduced word are detected by the distribution \(\mathbb{P}_{R,q}\). We prove the stronger statement that, for arbitrary reduced words \(R_1,R_2\), the equality \(\mathbb{P}_{R_1,q}=\mathbb{P}_{R_2,q}\) as functions of \(q\) holds if and only if \(R_1\) and \(R_2\) lie in the same commutation class. In particular, equality of distributions already forces the two words to represent the same permutation. The proof combines the Bott–Samelson interpretation with Demazure products, commutation-class invariants, and Hecke-algebraic arguments.

We study the relation between the Bogoliubov-de Gennes (BdG) equation and the Ginzburg-Landau (GL) equation for a BCS model without external fields. While previous rigorous derivations of GL theory from BCS theory have focused on energies and minimizers, here we consider arbitrary critical points in the relevant energy regime. For temperatures close to the critical temperature, we prove that every sufficiently small solution of the BdG equation admits an asymptotic factorization into a microscopic Cooper-pair profile and a macroscopic order parameter. The latter satisfies the GL equation up to an error that vanishes in the scaling limit. We also prove the opposite direction, going from a solution of the GL equation to an approximate solution of the BdG equation. Our analysis relies on a Birman-Schwinger reformulation of the BdG equation, a Lyapunov-Schmidt type reduction, and semiclassical estimates at low regularity.

We prove that the maximal dimension $d_N$ of an irreducible representation of the symmetric group $S_N$ satisfies $$d_N=\sqrt{N!} \, e^{-(\mathfrak{d}+o(1))\sqrt{N} }, \quad N\to \infty,$$ for some constant $\mathfrak{d}>0$. This answers a question raised by Vershik–Kerov in 1985.

An observation of J-P.~Serre implies that the generic monic cubic polynomial, unique among generic monic polynomials of degree at least two, has a root that is a power series in the discriminant; Serre asked for a formula. We give formulas that work over any field with an absolute value and in every characteristic. Over a complete non-archimedean field we identify the root intrinsically: it is the isolated root, the one farthest from the others. We also answer the next case of Serre's question, computing explicitly the distinguished ramified quadratic factor of the generic monic quartic. The methods combine Hensel's lemma, identities for generalized binomial series, and elementary non-archimedean analysis.

We introduce Fuzzy PyTorch, a framework for rapid evaluation of numerical variability in deep learning (DL) models. As DL is increasingly applied to diverse tasks, understanding variability from floating-point arithmetic is essential to ensure robust and reliable performance. Tools assessing such variability must be scalable, efficient, and integrate seamlessly with existing frameworks while minimizing code modifications. Fuzzy PyTorch enables this by integrating stochastic arithmetic into PyTorch through Probabilistic Rounding with Instruction Set Management, a novel library interfacing with Verificarlo, a numerical analysis compiler. The library offers stochastic rounding mode and a novel mode; up-down rounding. Comparative evaluations show Fuzzy PyTorch maintains model performance and achieves runtime reductions of 5x to 60x versus Verrou, a state-of-the-art tool. We further demonstrate scalability by running models from 1 to 341 million parameters, confirming applicability across small and large DL architectures. Overall, Fuzzy PyTorch provides an efficient, scalable, and practical solution for assessing numerical variability in deep learning, enabling researchers and practitioners to quantify and manage floating-point uncertainty without compromising performance or computational efficiency.

We study finite-time acoustic wave tracking using effective interior Dirac-type actuators generated by clusters of subwavelength resonant bubbles driven from exterior transducers. Starting from a time-domain point-interaction approximation, the full scattering problem is reduced to a finite-dimensional delayed system of interacting retarded monopoles. After localization to a bounded observation domain, we formulate the tracking problem on a prescribed finite-dimensional Dirichlet spectral subspace. At the ideal level, a full-rank spatial coupling condition yields exact tracking of arbitrary sufficiently smooth compatible trajectories in this subspace by interior point sources. These ideal actuators are realized using small, high-contrast bubbles, whose Laplace-domain transfer operator encodes the Minnaert resonances and collective attenuation. The principal mode of each cluster is isolated by the Perron–Frobenius structure and the intra-cluster spectral gap. Clusters tuned to the same band are then treated as a collective resonant block. The selected Dirichlet bands are matched with the corresponding Minnaert bands. For prescribed slowly time-modulated resonant pressure profiles, with sufficiently flat initialization, we construct causal exterior controls and obtain a quantitative tracking estimate on every fixed time interval $[0,T]$. In contrast, on frequency ranges uniformly separated from the Minnaert resonances, the actuator gain is $O(\varepsilon)$ and the corresponding realization cost is asymptotically unbounded as $\varepsilon \ll 1$, $\varepsilon$ being the maximum radius of the bubbles. Thus the Minnaert resonances convert accessible exterior excitations into robust effective interior controls.

We give a systematic construction of semiorthogonal decompositions of derived categories of coherent sheaves on quasi-smooth derived algebraic stacks over $\mathbb{C}$, where the summands are subcategories defined by weight conditions, and the inclusion functors are given by parabolic induction. The summands are indexed by the component lattice of the stack, a central combinatorial structure in intrinsic Donaldson-Thomas theory. As examples, we obtain semiorthogonal decompositions for moduli stacks of semistable $G$-bundles or $G$-Higgs bundles on a curve, and moduli stacks of de Rham or Betti $G$-local systems on a curve, for reductive groups $G$ not necessarily of type A.

Barcode entropy is an invariant of a Hamiltonian system – a Hamiltonian diffeomorphism or a Reeb flow – measuring its Morse or Floer theoretic complexity at a small scale. More specifically, it is the exponential growth rate of the number of not-too-short bars in the Floer or symplectic homology persistence module. Barcode entropy is closely related to topological entropy, even though they originate in different contexts, and in low dimensions they coincide. In these notes, we study barcode entropy and related invariants in various settings and explore their connections with pure dynamics features and, in particular, topological entropy. The methods build on techniques from symplectic topology and Floer theory, dynamical systems, and smooth integral geometry. We also touch upon some other applications of the machinery we develop. These notes are based on the mini-course given by the second author at the CIME summer school "Symplectic Dynamics and Topology" (Cetraro, Italy, June 16-20, 2025).

In this paper, we first introduce the definitions of random evolutionary system that associate with random evolutionary semigroup and the corresponding global weak or strong random attractor. Then we establish the existence result about global weak or strong random attractor and it's properties like invariance and weak or strong tracking features. Finally, we use our established results to the 3D stochastic Navier-Stokes equation with nonlinear colored noise.

We prove arithmeticity for two degree-six symplectic hypergeometric monodromy groups and thinness for a third case. These are the three open cases called C-32, C-47, and C-55 in the paper [BDN25] by Bajpai–Dona–Nitsche. This hence settles thinness of primitive integral symplectic hypergeometric monodromy groups of degree-six, according to the tabulation by Bajpai–Dona–Singh–Singh [BDSS21]. The arithmeticity certificates were found with AlphaEvolve and the thinness certificate is a ping pong table found by a harness that was configured to run ChatGPT 5.5. These certificates are then independently verified with exact matrix arithmetic over Q using a computer. We also include illustrations of limit sets of several degree-six symplectic hypergeometric monodromy groups.

Regression modeling of recurrent and terminal events continues to present methodological challenges in survival analysis. Existing approaches either make unverifiable assumptions about the dependency structure between the two event types or rely on the proportional intensity assumption for the marginal mean. A semiparametric regression model is proposed that is based on a novel weighted likelihood function, thereby targeting directly the marginal mean of the recurrent event. Our general model captures a large class of semiparametric regression models and accommodates external time-dependent covariate effects on the marginal mean intensity. We establish the consistency and asymptotic normality of the estimators and propose a sandwich estimator of the variance. We propose a novel simulation procedure that directly targets the marginal mean intensity of the recurrent events. In simulation studies, we demonstrate a strong performance of the weighted NPMLE under independent right-censoring. The practical utility of the proposed methodology is demonstrated through application to data from the STATCOPE trial, a large randomized clinical trial that investigated the efficacy of simvastatin for COPD exacerbations. We provide personalized predictions for the number of exacerbations and reassess the effect of simvastatin treatment, accounting for death as a competing terminal event for patients with GOLD stage 4.

This article investigates the exact controllability of three-dimensional stochastic Maxwell equations, a coupled system comprising two stochastic partial differential equations. The research establishes the observability inequality for the backward stochastic Maxwell equations using the multiplier method, and subsequently, proves the exact controllability of the forward equations. The control acting on the diffusion term is found to be indispensable, since exact controllability is destroyed when this control is removed; it is further proved that the controllability result obtained in this paper is achieved with a minimal number of controls. Finally, a numerical algorithm combining a central difference for spatial discretization, a midpoint scheme for temporal discretization and Lagrange multiplier method is proposed, yielding numerical results that offer the control value and lead to deeper insights into the underlying theoretical framework.

We consider a bilevel optimization problem in which the ground set is partitioned between two decision makers, a leader and a follower, whose optimization problems are interleaved. We study the Bilevel Independent Set problem, and its special case, the Bilevel Interval Selection problem, on different variants emerging from a combination of the type of leader's objective function, the type of follower's objective function, and the setting in which the follower reacts, i.e., either optimistically or pessimistically. Here we consider sum and bottleneck type objective functions. We investigate the computational complexity of all these variants for the Bilevel Independent Set problem, and sort them into their respective level of the polynomial hierarchy. Our results range from $\mathsf{P}$, $\mathsf{NP}$-completeness to $\Sigma_2^\mathsf{p}$-completeness. For the Bilevel Interval Selection problem, we give a dynamic programming algorithm running in time $\mathcal{O}(n^4\log n)$ for the variants in which the leader and the follower have objective functions of the sum type.

We show that if $T$ is a Dini-continuous Calderón–Zygmund operator satisfying $T(1)=0$, then the usual sparse domination for $T$ can be sharpened by replacing local averages with local mean oscillations. This extends a result of Benea and Bernicot for smoother kernels to the more general Dini-continuous setting. As an application, we characterize the Calderón–Zygmund operators for which a pointwise Sobolev-type inequality holds: this is the case if and only if $T(1)\in L^\infty$. This answers a recent question of Hoang, Moen and Pérez.

In this paper, we prove that every prime $p$ which is congruent to $4,7$ modulo $9$ is the sum of two rational cubes. This is $2/3$ of Sylvester's conjecture which has a history of nearly 150 years since 1879. In the proof, we use recent progress on Full BSD conjecture of rank $0$ elliptic curves in \cite{BF} to deduce that the Manin-Stevens constants of some families of elliptic curves are units. We also use recent solutions of Unbounded Denominators Conjecture in \cite{CDT} to prove that some cubic roots of modular functions are invariant under some congruence subgroups. Instead of using the Unbounded Denominators Conjecuture, we also give another conditional proof assuming the GRH for number fields or Artin's primitive root conjecture for arithmetic progressions.

We prove that, for every integer $r\ge 3$, the set $\Pi^{(r)}_\infty$ of Turán densities of (possibly infinite) families of $r$-graphs contains non-degenerate intervals, including an interval of the form $[1-\delta_r,1]$ for some $\delta_{r}>0$. This answers a question of Frankl, Peng, Rödl and Talbot from 2007. This also shows that the Hausdorff dimension of $\Pi^{(r)}_\infty$ has the maximum possible value 1, thus resolving a question of Grosu from 2016, whereas previously it was not even known whether it is non-zero. We also derive that the set of uniform Turán densities of finite families of $3$-graphs is dense in a non-degenerate interval.

In 1910, Élie Cartan famously realized the split real form of the exceptional Lie group $G_2$ as the symmetry group of the maximally symmetric rank 2 distribution on a 5-dimensional manifold with the small growth vector (2,3,5). In this paper, we discover a new appearance of $G_2$ in the geometric theory of distributions, arising from a rank 3 distribution on an 8-dimensional manifold with the growth vector $(3,6,8)$. The algebra of infinitesimal symmetries of this distribution at any point is 29-dimensional and isomorphic to $(\mathfrak{g}_2 \oplus \mathbb{R}) \ltimes W$, where $\mathfrak{g}_2$ is the Lie algebra of $G_2$ and $W$ is an adjoint module of $\mathfrak{g}_2$. Our model possesses three remarkable properties. First, it is maximally symmetric among all bracket-generating rank 3 distributions with a 6-dimensional square (a family that includes both (3,6,8) and (3,6,7,8) distributions). To the best of our knowledge, this is the first example of a family of distributions defined by a set of prescribed small growth vectors in which maximal symmetry is achieved by a member whose growth vector is not the longest. Second, this model provides the first counterexample to the conjecture that all bracket-generating rank 3 distributions with a 6-dimensional square are of maximal class at a generic point (which is known to hold in dimensions 6 and 7). Third, further analysis yields the control-theoretic consequence that all abnormal extremal trajectories of this model originating at any point of the ambient manifold have a corank of at least 2. To our knowledge, this is the first example with this property among bracket-generating distributions with generic small growth vector for a given rank and ambient dimension. We also give an interpretation of our model in terms of split-octonions, more precisely, in terms of a natural algebraic structure on the tangent bundle to split octonions.

Integral identities for Macdonald polynomials play an important role in modern mathematics and mathematical physics. Especially interesting are the Cherednik-Macdonald-Mehta (CMM) identities, with profound connections to Double Affine Hecke Algebras (DAHA) and representation theory of quantum groups. These identities are central in refined Chern-Simons theory, where they lead to refined S and T matrices and ultimately to refined knot invariants. We suggest an elliptic generalization of CMM identities, where trigonometric Vandermonde products are replaced by theta functions. At the same time Macdonald polynomials are promoted to Shiraishi functions – distinguished elliptic functions with several interesting avatars, from the non-stationary Ruijsenaars problem in integrable systems, to equivariant K-theory characters of the affine Laumon space in algebraic geometry, to surface defect partition functions in 5d super Yang-Mills theory. From the perspective of matrix models, we present an elliptic matrix model with a superintegrability property. We prove the suggested identities to the first order in the elliptic parameter.

Let $\mathbf{G}=\{G_1,\dots,G_{n-1}\}$ be a collection of not necessarily distinct $n$-vertex graphs with the same vertex set $V$. A path $P$ with $V(P)\subseteq V$ and $|E(P)|\leq n-1$ is called rainbow in $\mathbf{G}$, if there exists an injection $\phi\colon E(P)\to [n-1]$ such that $e\in E(G_{\phi(e)})$ for each $e\in E(P)$. The graph collection $\mathbf{G}$ is said to be rainbow panconnected if for every pair of vertices $x,y\in V$, there exists a rainbow path of $k$ vertices joining $x$ and $y$ in $\mathbf{G}$ for every integer $k\in \left[d_{\mathbf{G}}(x,y)+1, n\right]$, where $d_{\mathbf{G}}(x,y)$ is the length of a shortest rainbow path between $x$ and $y$ in $\mathbf{G}$. In this paper, we study the rainbow panconnectivity of $\mathbf{G}$ under the minimum degree condition. Our result improves upon the corresponding results of [J. Graph Theory, 104(2)(2023), 341–359] and [Electron. J. Combin., 32(4)(2025), \#P4.17].

For every fixed integer $t\geq 3$, we construct an $n$-vertex $K_{2,t+1}$-free graph containing $\Omega_t(n^2)$ copies of $K_{t,t}$. Combined with a simple counting argument, this shows that \[ \mathrm{ex}(n,K_{t,t},K_{2,t+1})=\Theta_t(n^2). \] This answers a question of Spiro.

Quantum graphs provide an analytically tractable setting for the study of Chern numbers and band degeneracies in periodic systems. We study the Chern numbers of energy bands in a two-dimensional square lattice quantum graph. We approach the problem by mapping the lattice to a single-vertex quantum graph with two loops of equal lengths pierced by magnetic fluxes. By establishing the degeneracy condition for its energy levels, we show that the model possesses two topological phases: a trivial phase, where the Chern numbers of all energy bands are $0$, and the nontrivial one, where the Chern numbers of successive energy bands alternate between $\pm1$. By applying the degeneracy condition, we calculate Chern-number phase diagrams analytically as a function of the node scattering matrix parameters and compare the results with numerical calculations.

We study rational double points over algebraically closed fields in arbitrary characteristics and completely classify the indecomposable objects in their singularity categories, which correspond to the vertices in their Auslander-Reiten quivers. Along the way, we present an alternative proof determining the configuration of these Auslander-Reiten quivers, and provide methods to handle the homotopy categories of matrix factorizations of isolated hypersurface singularities with computer algebra systems.

We study how an e-commerce firm should make real-time fulfillment decisions in a two-layer distribution network when multi-item customer orders arrive sequentially and future demand is unknown. The central managerial tension is whether to use scarce front distribution center (FDC) inventory to save current fulfillment cost or preserve that inventory for future orders that may be more valuable to serve locally. We formulate an adversarial online model with multiple FDCs, one regional distribution center (RDC), multi-unit multi-item orders, and item-specific and time-varying variable costs. Our theoretical objective is to characterize when simple, interpretable, and implementable fulfillment rules can perform nearly as well as an optimal clairvoyant planner. We develop a family of Gated Priority-based Greedy policies, derive competitive-ratio guarantees under both time-varying and time-invariant cost structures, and establish matching or near-matching lower bounds for any online algorithm. Numerical experiments show that the proposed policies perform strongly relative to generalized myopic and forecast-based benchmarks. The analysis yields managerial guidance on when local inventory should be protected, when splitting orders is worth the fixed-cost burden, and how the relative magnitudes of fixed and variable costs determine the value of more sophisticated optimization.

This work presents a novel family of well-balanced numerical schemes for hyperbolic systems of balance laws based on the kinetic relaxation approach. The method begins by transforming the original non-linear system into a linearized kinetic system with an increased number of variables. In this framework, non-linearities are shifted to the source term, and the connection to the macroscopic variables is maintained via a projection operator related to the Maxwellian equilibrium states. These relaxed systems are typically solved using splitting techniques, where the evolution is decomposed into two distinct steps: first, the transport stage solving the source-free kinetic system, and second, the relaxation stage governed by an ordinary differential equation where the time derivative is balanced by the kinetic source term. The primary objective is to design schemes that accurately preserve the steady-state solutions of the original macroscopic equations, noting that these do not necessarily coincide with the trivial equilibria of the relaxed system. This well-balanced property is achieved through an appropriate formulation of the kinetic system combined with consistent reconstruction operators. This approach ensures that the balance is maintained throughout both the transport stage and the projection stage, where the treatment of the macroscopic source term is carefully integrated. We demonstrate that this framework can be implemented using both Finite Volume (FV) and Semi-Lagrangian (SL) techniques in the transport step. We develop and analyze these schemes for the Finite Volume case in both explicit and implicit formulations. Results are validated through numerical benchmarks, including the Burgers equation with a source term, 1D Shallow Water Equations (SWE) with bathymetry preserving non-trivial moving water equilibria and the Euler equations under external potentials.

We consider the one dimensional space-periodic Vlasov-Poisson equations and construct, close to symmetric flat velocity strips, small amplitude traveling quasi-periodic electron-layers, namely strip-shaped patches of electrons in the phase space. These solutions are found for most values of the strip area. The proof uses a Nash-Moser construction together with reducibility arguments based on pseudo-differential homogeneous expansions and KAM reductions. Thanks to a suitable linear transformation of the unknowns, we reveal a connection between the electron patch problem and the classical, physically relevant, electronic Euler-Poisson system with cubic pressure law. As a direct, non trivial consequence, we obtain small-amplitude quasi-periodic traveling waves for this model as well. To the best of our knowledge, this work provides the first rigorous example of construction of quasi-periodic solutions for kinetic models.

Let $q$ be an odd prime power and let $\F_q$ be the finite field with $q$ elements. Let $\mathcal{P}(n)$ be the set of monic irreducible polynomials of degree $n$ over $\mathbb{F}_q$. For $f=t^n+f_{n-1}t^{n-1}+\cdots+f_0\in\mathcal{P}(n)$, fix coefficients $c_0,\ldots,c_m\in\mathbb{F}_q$ with $c_m\ne0$ and put $$ Q_A(f)=\sum_{j=0}^m c_j\sum_{i=j}^n f_i f_{i-j}+\ell_n(f),$$ where $\ell_n$ is an arbitrary linear form in the coefficients of $f$ and $f_n=1$. We prove that $Q_A$ is equidistributed on $\mathcal{P}(n)$: for every $\gamma\in\mathbb{F}_q$, $$\#\{f\in\mathcal{P}(n):Q_A(f)=\gamma\}=\frac{\#\mathcal{P}(n)}{q}+O_A(q^{19n/20+o(n)}),$$ as \(n\to\infty\), with $q$ and the quadratic band fixed. This extends the finite-field Rudin–Shapiro result from nearest-neighbour correlations to arbitrary fixed symmetric Laurent symbols. The proof combines Vaughan's identity with rank estimates for Toeplitz forms; the main new ingredient is an averaged rank-defect estimate for reciprocal symbols in the central Type I range.

The aim of this Lecture Note is to introduce the Signal Processing (SP) community to a powerful yet still under-utilised tool: the semiparametric statistics. In short, the semiparametric framework allows us to estimate or perform hypothesis testing on a finite-dimensional parameter $\boldsymbol{\theta} \in \Theta \subseteq \mathbb{R}^p$ in the presence of an infinite-dimensional nuisance parameter $g \in \mathcal{S}$ (i.e. a function), such as the density of the noise. Clearly, this framework is general enough to include almost every SP application. Remarkably, as the title suggests drawing on George R. R. Martin's famous book series, the greatest advantage of semiparametric statistics over parametric and non-parametric ones lies in the fact that it is able to reconcile two seemingly dichotomous concepts: statistical efficiency and distributional robustness. To explain exactly what this means, in this Lecture Note we will focus our attention on the famous and fundamental symmetric location problem.