Peer Revenue Play credits. Real papers.

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Mathematics

math Watch Mathematics (sign in to use)

Wednesday 8 April

Probability

Some probabilistic properties and time-changed versions of a renewal process based on Mittag-Leffler waiting times

In this paper, we obtain some additional probabilistic properties of the renewal process $\{\hat{N}_{\alpha}(t)\}_{t\ge0}$, $0<\alpha\le 1$ introduced by Beghin and Orsingher (2010). A time-changed relationship connecting $\{\hat{N}_{\alpha}(t)\}_{t\ge0}$ with its special case $\{\hat{N}(t)\}_{t\ge0}$ by means of the random time process $\{T_{2\alpha}(t)\}_{t>0}$ whose distribution is related to a fractional diffusion equation is established. We compute its various distributional properties such as the variance, factorial moments, moment generating function, moments, covariance in the Laplace domain, etc. We show that the ratios given by $\{\hat{N}_{\alpha}(t)\}_{ t \ge 0}$ and its power over their means tend to $1$ in probability. Moreover, we derive an integral form of its bivariate distribution and describe the scaling limits of its marginal distributions. It is also shown that its one-dimensional distributions are not infinitely divisible. Furthermore, we study the compound version of $\{\hat{N}_{\alpha}(t)\}_{ t \ge 0}$ and discuss an application to ruin theory. Later, we consider two time-changed versions of $\{\hat{N}_{\alpha}(t)\}_{ t \ge 0}$ which are obtained by time-changing it with an independent Lévy subordinator and its inverse. Some distributional properties and examples are discussed for these time-changed processes.
Read the full abstract

Algebraic Geometry

Reconstruction of F-cohomological field theories on moduli of compact type

We prove an analogue of Givental-Teleman reconstruction for F-cohomological field theories on the moduli space of compact type. We apply it to reconstruct the restriction of the extended $r$-spin classes to the extended direction and deduce relations between $\kappa$-classes (both in compact type).
Read the full abstract

Algebraic Geometry

Stable degeneration and birational geometry

This expository article is based on the author's talk at the Kinosaki Algebraic Geometry Symposium 2025. We discuss some recent progress surrounding stable degeneration in algebraic K-stability theory.

Algebraic Geometry

A note on b-divisors and filtrations on a local ring

In this note, we prove a correspondence between filtrations and b-divisors over a general class of Noetherian local domains. As an application in the global setting, we prove a recent conjecture of Roé-Urbinati.

Analysis of PDEs

Compactness of Solutions to Sub-Elliptic Equations with Potential on the Heisenberg Group

In this paper, we investigate the compactness of nonnegative solutions to a critical sub-elliptic equation with a nonnegative potential on the Heisenberg group. We establish compactness in all dimensions. Moreover, we show that if a sequence of solutions blows up, both the potential and its sub-Laplacian must vanish at the blow-up point. Our analysis overcomes the inherent geometric and analytical challenges posed by the Heisenberg group, including the degeneracy of the sub-Laplacian, its non-commutative structure, and the anisotropic dilation symmetry.
Read the full abstract

All authors Jiechen Qiang, Zhongwei Tang, Yichen Zhang, Ning Zhou

Group Theory

Sha-rigidity of adjoint Chevalley groups of types $A_1$, $A_2$, $B_2$, $G_2$ over commutative rings

We prove that every locally inner (class-preserving) endomorphism of adjoint Chevalley groups and their elementary subgroups over commutative rings is inner for the root systems A1, A2, B2 (assuming 2 is invertible in the ring), and for G2 (assuming 2 and 3 are invertible). As a consequence, these groups are Sha-rigid. The proofs are direct and do not rely on classification of automorphisms or structural results about injective endomorphisms.
Read the full abstract

Category Theory · Cross-listed from cs.LG

Weaves, Wires, and Morphisms: Formalizing and Implementing the Algebra of Deep Learning

Despite deep learning models running well-defined mathematical functions, we lack a formal mathematical framework for describing model architectures. Ad-hoc notation, diagrams, and pseudocode poorly handle nonlinear broadcasting and the relationship between individual components and composed models. This paper introduces a categorical framework for deep learning models that formalizes broadcasting through the novel axis-stride and array-broadcasted categories. This allows the mathematical function underlying architectures to be precisely expressed and manipulated in a compositional manner. These mathematical definitions are translated into human manageable diagrams and machine manageable data structures. We provide a mirrored implementation in Python (pyncd) and TypeScript (tsncd) to show the universal aspect of our framework, along with features including algebraic construction, graph conversion, PyTorch compilation and diagram rendering. This lays the foundation for a systematic, formal approach to deep learning model design and analysis.
Read the full abstract

Optimization and Control

Non-Lipschitz Inertial Contraction-Type Method for Monotone Variational Inclusion problems

This study explores an inertial-based contraction-type approach for addressing monotone variational inclusion problems (in short, MVIP) within real Hilbert spaces. Most contraction-type techniques assume Lipschitz continuity and monotonicity or co-coercivity (inverse strongly monotone) of the single-valued operator. However, the key advantage of the proposed method is that it does not rely on the coercivity condition and the Lipschitz continuity for the single-valued operator. A weak convergence result has been achieved for the proposed algorithm with a convergence rate $\mathcal{O}\left(1/\sqrt{k}\right)$. In addition, the maximal and strong monotonicity of the set-valued operator is used to establish a strong convergence result with the linear convergence rate. To demonstrate the effectiveness of our proposed method, we conduct numerical experiments focused on signal recovery problems.
Read the full abstract

All authors Feeroz Babu, Syed Shakaib Irfan, Jen-Chih Yao, Xiaopeng Zhao

Information Theory · Cross-listed from cs.CL

Efficient Learned Data Compression via Dual-Stream Feature Decoupling

While Learned Data Compression (LDC) has achieved superior compression ratios, balancing precise probability modeling with system efficiency remains challenging. Crucially, uniform single-stream architectures struggle to simultaneously capture micro-syntactic and macro-semantic features, necessitating deep serial stacking that exacerbates latency. Compounding this, heterogeneous systems are constrained by device speed mismatches, where throughput is capped by Amdahl's Law due to serial processing. To this end, we propose a Dual-Stream Multi-Scale Decoupler that disentangles local and global contexts to replace deep serial processing with shallow parallel streams, and incorporate a Hierarchical Gated Refiner for adaptive feature refinement and precise probability modeling. Furthermore, we design a Concurrent Stream-Parallel Pipeline, which overcomes systemic bottlenecks to achieve full-pipeline parallelism. Extensive experiments demonstrate that our method achieves state-of-the-art performance in both compression ratio and throughput, while maintaining the lowest latency and memory usage. The code is available at https://github.com/huidong-ma/FADE.
Read the full abstract

All authors Huidong Ma, Xinyan Shi, Hui Sun, Xiaofei Yue, Xiaoguang Liu, Gang Wang, Wentong Cai

Operator Algebras

Generalised diagonal dimension and applications to large-scale geometry

In this paper, we introduce a generalised diagonal dimension. We explain why the generalised diagonal dimension extends the notion of diagonal dimension defined by Li, Liao, and Winter, and under which conditions these dimensions coincide. We prove permanence properties for the generalised diagonal dimension and compare it with the nuclear dimension. We investigate applications of the generalised diagonal dimension in large-scale geometry; specifically, we show that the generalised diagonal dimension of a noncommutative Cartan subalgebra in the C*-algebra of finite-propagation operators on a uniformly locally finite metric space is equal to the asymptotic dimension of the space.
Read the full abstract

Information Theory

The Random Subsequence Model and Uniform Codes for the Deletion Channel

We introduce the Random Subsequence Model, a spin glass model on pairs of random strings $(X,Y) \in \{0,1\}^N \times \{0,1\}^M$ whose partition function counts subsequence embeddings of $Y$ into $X$. We study two variants: the null model, where $X$ and $Y$ are independent and uniform, and the planted model, where $X$ is uniform and $Y$ is a uniformly-random length-$M$ subsequence of $X$. We connect the Random Subsequence Model to longstanding problems in various fields, including the best rate achievable by uniformly-random codes in the deletion channel, the longest common subsequence problem between two random strings, and models of directed polymers in statistical physics. In the regime where $N,M\to\infty$ at a fixed ratio $\alpha = M/N \in (0,1)$, we exhibit strict asymptotic separations between the null annealed free energy and the quenched free energies of the null and planted models at all values of the density parameter $\alpha$. This suggests that these models are in a spin glass phase at zero temperature throughout the entire dense regime. As a consequence, we show that uniformly-random codes achieve a positive rate in the deletion channel for all deletion probabilities $p\in [0,1),$ settling multiple conjectures of the second author, Isik and Weissman (2024) and proving the first such positive rate result for the regime $p \geq 1/2$. We also give an exact analytic formula for the annealed free energy of the planted model for all values of the density parameter. This implies a corresponding analytic upper bound on the best rate achievable by uniformly-random codes in the deletion channel, complementing the lower bound from our first result. Our upper and lower bounds for the capacity of the deletion channel under uniform codes are far closer to each other than the best known upper and lower bounds for the capacity of the deletion channel.
Read the full abstract

Dynamical Systems · Cross-listed from physics.soc-ph

Emergence of cooperation in nonlinear higher-order public goods games

Evolutionary game theory has provided substantial contributions to explain the emergence of cooperation under unfavorable conditions in ecology, economics, and the social sciences. Recently, inspired by newly available empirical evidence on group interactions, higher-order networks have emerged as a natural framework to encode multiplayer games in structured populations. Here, we study the emergence of cooperation in nonlinear public goods games on hypergraphs, where collective reinforcement captures the synergistic or discounting effect associated with each additional cooperator. In well-mixed populations, when all games have the same number of players, the system displays a transition whose nature changes from continuous to discontinuous depending on the form of nonlinearity. By contrast, when games involving different numbers of players coexist, an additional bistability region between an active coexistence state and full cooperation may emerge. We further find that scale-free hypergraphs promote cooperation, highlighting the crucial role played by both the initial placement of cooperators and the presence of hyperdegree correlations. Overall, our results provide a comprehensive characterization of nonlinear public goods games on hypergraphs and open avenues for richer models of evolutionary dynamics of multiplayer games on structured populations.
Read the full abstract

All authors Jaume Llabrés, Onkar Sadekar, Federico Malizia, Federico Battiston

Probability

Transition probabilities of step-reinforced random walks

The step-reinforced random walk (SRRW) either repeats a uniformly chosen past step or takes a fresh independent step. We consider a generalized SRRW on groups, where the selected past step is transformed by a random map. The transformations are independent of the fresh steps but may be dependent on one another. For every reinforcement parameter $\alpha<1$, we obtain upper bounds on transition probabilities in terms of the geometry of the group. On $\mathbb{R}^d$, a non-singular step distribution yields bounds of order $n^{-d/2}$ for balls of any fixed radius, uniformly over their centers and without any moment assumption, and hence transience for $d\geq 3$. On finitely generated groups, we give bounds in terms of the isoperimetric profile. We also prove exponential decay for the elephant random walk on every nonamenable Cayley graph with a finite symmetric generating set and every memory parameter $p<1$. In particular, this answers a question of Mukherjee for Cayley trees.
Read the full abstract

Optimization and Control

A Trajectory-Based Approach to Controlled Invariance and Recursively Feasible MPC

In this paper, we revisit the computation of controlled invariant sets for linear discrete-time systems through a trajectory-based viewpoint. We begin by introducing the notion of convex feasible points, which provides a new characterization of controlled invariance using finitely long state trajectories. We further show that combining this notion with the classical backward fixed-point algorithm allows for the computation of the maximal controlled invariant set. Building on these results, we propose a model predictive control (MPC) scheme that guarantees recursive feasibility without relying on precomputed terminal sets. Finally, we formulate the search for convex feasible points as an optimization problem, yielding a practical computational method for constructing controlled invariant sets. The effectiveness of the approach is illustrated through numerical examples.
Read the full abstract

Information Theory

Robust Hybrid Beamforming with Liquid Crystal Antennas and Liquid Neural Networks

Sub-terahertz (sub-THz) multi-user multiple-input multiple-output (MU-MIMO) systems unlock immense bandwidth for 6G wireless communications. However, practical deployment of wireless systems in sub-THz bands faces critical challenges such as increased atmospheric absorption, reduced channel coherence time due to increased Doppler spread at higher carrier frequencies, and hardware bottlenecks as low-loss sub-THz phase shifters are difficult to realize. To overcome the hardware and channel estimation challenges of sub-THz systems, this paper proposes a hybrid beamforming (BF) framework that integrates reconfigurable liquid crystal (LC) antennas with a liquid neural network (LNN) for transmitter. Specifically, we employ an LC antenna as the analog BF stage of a hybrid BF architecture, exploiting its voltage-driven permittivity tunability to achieve high-gain beam steering without the need for lossy phase shifters. For digital BF, we utilize an ordinary differential equations-defined LNN to learn temporal channel dynamics, and use a manifold optimization technique to compress the search space. We validated the proposed method on simulated site-specific 108 GHz ray-tracing channels in an urban scenario using NYURay, a ray-tracing simulator validated against 142 GHz propagation measurements. The 108 GHz carrier frequency matches the operating band of the LC antenna hardware. The proposed method achieves an 88.6% spectral efficiency (SE) gain and higher robustness to imperfect channel estimation compared to the learning-aided gradient descent and gated recurrent unit machine learning baselines, and 1.9 times higher SE than the 3GPP TR~38.901 standard antenna model, highlighting the potential of LC-based hardware for sub-THz communications.
Read the full abstract

All authors Xinquan Wang, Mingjun Ying, Hongren Chen, Guanyue Qian, Xingchen Liu, Peijie Ma, Dipankar Shakya, Christos Argyropoulos, Theodore S. Rappaport

Algebraic Geometry

A note on Bondal's conjecture

We prove that the connection vector fields associated to ample Poisson line bundles are not locally hamiltonian unless the Poisson structure is zero. We use this result to provide further evidence on Bondal's conjecture regarding the dimensions of the degeneracy loci of a holomorphic Fano Poisson manifold.
Read the full abstract

Optimization and Control

An Inexact Trust-Region Method for Structured Nonsmooth Optimization with Application to Risk-Averse Stochastic Programming

We develop a trust-region method for efficiently minimizing the sum of a smooth function, a nonsmooth convex function, and the composition of a convex Lipschitz continuous function with a smooth function. Optimization problems with this structure arise in numerous applications including risk-averse stochastic programming and subproblems for nonsmooth penalty nonlinear programming methods. Our algorithm permits the use of inexact value and derivative information, enabling the solution of infinite-dimensional problems governed by, e.g., partial differential equations (PDEs). We prove global convergence of our method and under additional regularity assumptions, demonstrate that the sequence of iterates accumulates at a stationary point of our target problem. We demonstrate our method's efficiency on two PDE-constrained optimization examples, showing that its performance is invariant to the PDE discretization size.
Read the full abstract

Complex Variables

On weak Wolff–Denjoy theorem for certain non-convex domains

In this paper, we provide a class of domains in $\mathbb{C}^3$, such that every holomorphic self-map of that domain either has a fixed point or the sequence of iterates is compactly divergent. In particular, it follows that the symmetrized bidisc, symmetrized tridisc, tetrablock, pentablock are in the aforementioned class of domains. We also prove that the fixed point set of a holomorphic self map of symmetrized bidisc and tetrablock is either empty set or a holomorphic retract. For the symmetrized bidisc, given a holomorphic self-map such that the sequence of iterates is compactly divergent, we also provide a description of its target set.
Read the full abstract

Numerical Analysis · Cross-listed from quant-ph

Quantum Gibbs sampling through the detectability lemma

Gibbs state preparation is an important subroutine in quantum computing. In this work we use the detectability lemma to improve Gibbs state preparation. Specifically, we design new Gibbs state preparation methods that do not rely on simulating Lindbladian evolution, thus avoiding the overhead from it. For local Lindbladians consisting of $M$ terms, this approach reduces the cost by a factor of $O(M)$. We also combine the detectability lemma operator and quantum singular value transformation to implement ground state projection operators of frustration-free Hamiltonians, resulting in a quadratic speedup in the spectral gap dependence. Applying this method to Lindbladians for the Gibbs state of local commuting Hamiltonians, we achieve quadratically better dependence on the Lindbladian spectral gap.
Read the full abstract

All authors Di Fang, Jianfeng Lu, Yu Tong, Chu Zhao

Probability · Cross-listed from cs.LG

Diffusion Processes on Implicit Manifolds

High-dimensional data are often assumed to lie on lower-dimensional manifolds. We study how to construct diffusion processes on this data manifold using only point cloud samples and without access to charts, projections, or other geometric primitives. Here, we introduce Implicit Manifold-valued Diffusions (IMDs), a data-driven mathematical formalism for defining stochastic differential equations in the original high-dimensional space that describe drifting Brownian particles evolving intrinsically on the underlying manifold. Our construction hinges on approximating the corresponding infinitesimal generator of the diffusion process using a proximity graph over the data and using the carré-du-champ of the generator, which encodes the local tangent spaces of the manifold and lifts the intrinsic process into ambient coordinates. We show that as the number of samples grows, our discrete diffusion process converges in law on the space of probability paths to its smooth manifold counterpart. We further present an Euler-Maruyama scheme for the numerical integration of IMDs. We validate our framework using numerical experiments on synthetic manifolds and the MNIST data manifold, showing that IMDs remain confined over the manifold and enable its guided exploration. Our work provides the mathematical foundation and practical implementations of diffusion processes on data manifolds, opening new avenues for manifold-aware sampling, exploration, and generative modeling.
Read the full abstract

All authors Victor Kawasaki-Borruat, Clara Grotehans, Pierre Vandergheynst, Adam Gosztolai

Probability

Mixing times of step-reinforced random walks

We study the mixing time of a non-Markovian process, the step-reinforced random walk (SRRW) on a finite group. This process differs from a classical random walk in that at each integer time, with probability $\alpha$ the next step is chosen uniformly from the previous steps of the walk. We prove that the distribution of the SRRW converges to the uniform distribution exponentially fast if the walk is irreducible and aperiodic. When the step distribution is either symmetric, a class function, or has an atom at the identity, we relate the mixing time of the SRRW to the spectral gap and the mixing time of the underlying walk. For the reinforced (lazy) simple random walk on $L$-cycles, we show that the mixing time undergoes a phase transition at $\alpha=1/2$ and the reinforcement reduces the mixing time to order $L^{1/\alpha}$ for $\alpha>1/2$. On the $d$-dimensional hypercube, the reinforcement slows down mixing, and the SRRW exhibits cutoff as $d\to\infty$ at time $d\log(d)/[F(\alpha)(1-\alpha)]$, with an $O(d)$ window for each fixed $\alpha\in(0,1)$, where $F(\cdot)$ is a hypergeometric function.
Read the full abstract

Numerical Analysis

Robust preconditioning for an HDG discretization of the time-dependent Stokes equations

We present parameter-robust preconditioners for linear systems that arise after applying static condensation to a hybridizable discontinuous Galerkin (HDG) discretization of the time-dependent Stokes problem. Building upon the theoretical framework introduced in our previous work [SIAM Journal on Scientific Computing, 47(6):A3212-A3238, 2025], we extend the analysis to derive new preconditioners that remain robust with respect to all physical and discretization parameters. The construction relies on first establishing uniform well-posedness of the HDG formulation (before static condensation) through appropriately defined norms. Based on this result, we identify sufficient conditions that a norm on the face space must satisfy to guarantee parameter-robustness of the resulting preconditioner for the statically condensed HDG system. Numerical experiments in two and three dimensions verify our theoretical results.
Read the full abstract

Number Theory

Weighted averages of arithmetic functions and applications to equidistribution and ergodic theory

For a wide range of functions $W\colon\mathbb{N}\to\mathbb{N}$, we establish a general result for estimating weighted averages of the form\[\mathbb{E}^{W}_{n \le N} f(\vartheta(n))= \frac{1}{W(N)}\sum_{n=1}^N (W(n)-W(n-1))f(\vartheta(n)),\]where $f\colon \{1,\ldots,N\}\to\mathbb{C}$ is an arbitrary function, and $\vartheta(n)$ is any arithmetic function that adheres to a certain Gaussian distribution condition. (For instance, one may take $\vartheta(n)=\Omega(n)$, where $\Omega(n)$ counts the number of prime factors of $n$ with multiplicity, or $\vartheta(n)=s_q(p_n)$, where $s_q$ is the sum-of-digits function in base $q$ and $p_n$ denotes the $n$-th prime. Additional natural examples are discussed in the paper.) Building on our main theorem, we show that if $h(n)$ is a function from a Hardy field with polynomial growth then $(h(\vartheta(n)))_{n\in\mathbb{N}}$ is uniformly distributed mod $1$ if and only if one of the following (mutually exclusive) conditions is satisfied: (i) $\lim_{x\to\infty} \frac{|h(x)-p(x)|}{x \log x}=\infty$ for all $p(x)\in \mathbb{Q}[x]$; (ii) $\lim_{x\to\infty}\frac{|h(x)-p(x)|}{\sqrt{x}}=\infty$ for each $p(x)\in \mathbb{Q}[x]$ and there exists $q(x)\in \mathbb{Q}[x]$ such that $\lim_{x\to\infty}\frac{|h(x)-q(x)|}{x}<\infty$. This leads to several novel applications. For example, it follows that $(\Omega(n)^c)_{n\in\mathbb{N}}$ is uniformly distributed mod $1$ if and only if $c$ is a non-integer greater than $\frac{1}{2}$.
Read the full abstract

Probability

Stopping on the last success with unknown odds: asymptotic minimax optimality of the plug-in rule

We study the last-success problem for sequential Bernoulli trials in the homogeneous setting where $X_1,\ldots,X_n$ are i.i.d. $\mathrm{Bernoulli}(p)$ but the success probability $p\in(0,1)$ is unknown to the decision maker. When $p$ is known, Bruss' sum-the-odds theorem yields an optimal threshold rule with win probability $V_n(p)$; when it is not, the odds driving this threshold must be learned from the very sequence on which one is trying to stop, which turns the problem into a genuinely statistical decision problem over the class of $p$-blind rules—those depending on the data but not on $p$. Writing $W_n^\pi(p)$ for the win probability of such a rule, we show that, for any $p_0\in(0,\tfrac12)$, $$ \lim_{n\to\infty}\sqrt n\,\inf_\pi\sup_{p\in[p_0,1)}\bigl(V_n(p)-W_n^\pi(p)\bigr) = C_\star , $$ with $C_\star:=\tfrac12\sup_{u>0}u\Phi(-u)\approx0.085$ (here, $\Phi$ is the standard normal distribution function), and that this exact constant is attained by the natural plug-in odds rule, which is therefore asymptotically minimax optimal. The result is in fact local: at every transition point $1/k$ of the oracle threshold, the deficit admits an exact local minimax constant proportional to $\gamma_k=\tfrac{1}{\sqrt k}(1-\tfrac1k)^{k-3/2}$, and $C_\star$ is the largest of these, attained at $k=2$. We further quantify the cost of the natural sample-splitting alternative, show that the plug-in rule is asymptotically oracle-optimal in the sparse regime $p=p_n\to0$ with $np_n\to\infty$, and prove that no $p$-blind sequence of rules, even randomized, can converge to the oracle uniformly over $p\in(0,1)$, the obstruction being located in the critical window where $p$ is of order $1/n$.
Read the full abstract

Numerical Analysis

A spectral method for the rapid evaluation of hyperbolic potentials in two dimensions using windowed Fourier projection

We present a fast algorithm for evaluating the (non-smooth) solution of the free-space two-dimensional (2D) scalar wave equation with many point sources, each with a high-frequency band-limited time signature. Such an algorithm is key to an efficient time-domain scattering solver using spatially-discretized hyperbolic layer potentials. Given $M$ sources/targets and $N_t$ time steps, direct evaluation costs $O(M^2N_t^2)$, due to the history dependence. We develop a quasi-linear scaling algorithm that splits the solution at a given time into (a) a non-smooth time-local part, (b) a (smooth) near history involving sources up to ${\mathcal O}(1)$ domain traversal times into the past, plus (c) a (very smooth) far history comprising all waves emitted before the near history. The local part is computed directly via high-order quadrature. A naive spatial Fourier transform for (b) plus (c) would be both slowly converging and arbitrarily oscillatory as time progresses. Yet in (b) the oscillations are controlled, so we use the recent truncated windowed Fourier projection (TK-WFP) method to give rapid convergence. For (c) – present due to the weak Huygens' principle – we exploit a new large-time sum-of-exponentials approximation of the free-space wave kernel. Numerical examples with up to a million sources and targets, a domain of $300\times 300$ wavelengths, and 6-digit accuracy, show an acceleration of five orders of magnitude relative to direct evaluation.
Read the full abstract

Numerical Analysis · Cross-listed from stat.ML

Amortized Filtering and Smoothing with Conditional Normalizing Flows

Bayesian filtering and smoothing are central to data assimilation in nonlinear dynamical systems. Recent advances in deep generative models provide flexible approximations of the associated non-Gaussian posterior distributions. However, several existing approaches require the score to be re-estimated or a transport map to be constructed at each assimilation step. We propose an amortized framework for filtering and smoothing that reuses trained conditional models across observation sequences and assimilation times. The framework jointly learns a shared recurrent summary network and two conditional normalizing flows from simulated state and observation trajectories. The recurrent network represents each observation history by a fixed-dimensional summary that conditions the filtering approximation and, together with the next state, the backward transition approximation. An information-theoretic analysis shows that, under the Markov assumption, a summary sufficient for filtering is also sufficient for the backward kernel. Combining the terminal filtering approximation with the backward kernels yields an approximate joint smoothing distribution. Numerical experiments on advection-diffusion, Burgers, and Lorenz systems demonstrate the accuracy of the proposed filtering and smoothing approximations and characterize the evolution of their errors beyond the training horizon.
Read the full abstract

All authors Tiangang Cui, Xiaodong Feng, Chenlong Pei, Xiaoliang Wan, Tao Zhou

Analysis of PDEs

Geometrically defined asymptotic coordinates in General Relativity

We review and announce recent results on the asymptotic behavior of asymptotically Euclidean relativistic initial data sets and asymptotic foliations thereof. In particular, we discuss the geometrization of asymptotic flatness and of asymptotic geometric (in-)\-variants such as mass, energy, linear momentum, angular momentum and center of mass as well as their relations to certain geometric asymptotic foliations such as the CMC- and STCMC-foliations.
Read the full abstract

Optimization and Control

Model-Free Aggregative Cooperative Optimization via Randomized Gradient-Free Minimization and Exploration Momentum

Aggregative cooperative optimization problems arise in distributed decision-making settings where each agent's objective depends on its own decision as well as on an aggregate variable capturing global system behavior. Motivated by practical scenarios where gradient information is unavailable, this paper introduces a randomized gradient-free algorithm, named ARGFree, for solving such problems. ARGFree combines finite-difference gradient approximations with a set of tracking variables, emulating the behavior of a gradient-based method. We prove that ARGFree converges in expectation to an approximate optimizer, with the approximation error stemming from the use of a randomized gradient estimator. To enhance performance in high-dimensional settings, we further propose an improved variant, ARGFree-EM, which incorporates momentum in the exploration signals to smooth sudden fluctuations in the gradient exploration signals and thereby improve the accuracy of the underlying distributed tracking mechanism. To the best of our knowledge, the class of ARGFree methods is the first in the literature capable of solving aggregating cooperative optimization problems without gradient information.
Read the full abstract

All authors Amir Mehrnoosh, Giuseppe Speciale, Riccardo Brumali, Giuseppe Notarstefano, Gianluca Bianchin

Numerical Analysis

Cut Finite Element Methods for Convection-Diffusion in Mixed-Dimensional Domains

We develop a cut finite element method (CutFEM) for convection–diffusion problems posed on mixed-dimensional domains, i.e., unions of manifolds of different dimensions arranged in a hierarchical structure where lower-dimensional components form parts of the boundaries of higher-dimensional ones. Such domains arise, for instance, in the modeling of fractured porous media with intersecting fractures. The model problem is formulated in a compact abstract form using mixed-dimensional directional derivative and divergence operators, which allows the problem to be expressed in a way that closely resembles the classical convection–diffusion equation. The proposed CutFEM is based on a fixed background mesh that does not conform to the geometry, with each manifold component represented through its associated active mesh. The method employs continuous piecewise linear elements together with weak enforcement of coupling conditions and suitable stabilization. We prove a priori energy norm error estimates under a global uniform-diffusion assumption, with corresponding extensions to solutions of reduced regularity $u\in H^s$, $1\le s<2$, and derive conditional estimates for the globally pure-convection case. Partially degenerate configurations, in which diffusion is present only on selected components, are explored numerically. The experiments report convergence in both the energy and $L^2$-norms and illustrate the performance of the method.
Read the full abstract

All authors Erik Burman, Peter Hansbo, Mats G. Larson, Karl Larsson, Shantiram Mahata

Numerical Analysis

Deterministic sketching for Krylov subspace methods

Randomized sketching is currently introduced into every area of numerical linear algebra. In Krylov subspace methods, it allows runtime savings at the cost of small accuracy reductions. This work offers a different view on sketching in Krylov subspace methods by analyzing what subspace embeddings are obtained by arbitrary sketching matrices. The analysis gives rise to a deterministic sketching approach leveraging row subset selection techniques that yield subspace embeddings holding with probability 1. We propose deterministically sketched Krylov subspace methods for matrix functions, linear systems, and eigenvalue problems that show a similar performance to their randomly sketched counterparts.
Read the full abstract