Peer Revenue Equity in science

Papers

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

Mathematics

math Watch Mathematics (sign in to use)

Thursday 5 March

Statistics Theory · Cross-listed from q-fin.ST

Asymptotic Separability of Diffusion and Jump Components in High-Frequency CIR and CKLS Models

This paper develops a robust parametric framework for jump detection in discretely observed CKLS-type jump-diffusion processes with high-frequency asymptotics, based on the minimum density power divergence estimator (MDPDE). The methodology exploits the intrinsic asymptotic scale separation between diffusion increments, which decay at rate $\sqrt{\Delta_n}$, and jump increments, which remain of non-vanishing stochastic magnitude. Using robust MDPDE-based estimators of the drift and diffusion coefficients, we construct standardized residuals whose extremal behavior provides a principled basis for statistical discrimination between continuous and discontinuous components. We establish that, over diffusion intervals, the maximum of the normalized residuals converges to the Gumbel extreme-value distribution, yielding an explicit and asymptotically valid detection threshold. Building on this result, we prove classification consistency of the proposed robust detection procedure: the probability of correctly identifying all jump and diffusion increments converges to one under proper asymptotics. The MDPDE-based normalization attenuates the influence of atypical increments and stabilizes the detection boundary in the presence of discontinuities. Simulation results confirm that robustness improves finite-sample stability and reduces spurious detections without compromising asymptotic validity. The proposed methodology provides a theoretically rigorous and practically resilient robust approach to jump identification in high-frequency stochastic systems.
Read the full abstract

Optimization and Control

Solution of a bilevel optimistic scheduling problem on parallel machines

We consider the uniform parallel machines scheduling problem in the context of optimistic bilevel optimization, where two speed options are considered. In this scenario, the leader aims to minimize the weighted number of tardy jobs, while the follower seeks to minimize the total completion time on a set of uniform machines. This problem has practical applications in Industry 4.0. We show that this problem is NP-hard in the strong sense by providing a reduction from the Numerical 3-Dimensional Matching problem and we provide a moderately exponential-time dynamic programming algorithm. The problem is solved by means of a concise MIP formulation and a branch-and-bound algorithm that embeds a column generation approach for the lower bound computation. Computational experiments are presented for instances with up to 80 jobs and 4 machines while larger problems are out of reach for the proposed approaches.
Read the full abstract

All authors Quentin Schau, Olivier Ploton, Vincent T'kindt, Han Hoogeveen, Federico Della Croce, Jippe Hoogeveen

Algebraic Geometry

Bruhat-Tits group schemes over higher dimensional base-II

We prove that split reductive BT group schemes over a higher dimensional base are affine. Our method also gives a new construction of higher BT-group schemes more general than parahoric ones. The new ingredients are an extension of J.-K.Yu's construction in \cite{yu} to higher dimensional bases, Néron-Raynaud dilatations of subgroup schemes on divisors, combined with techniques from \cite{bt2} and the structure theory developed in \cite{bp}.
Read the full abstract

Combinatorics

Minimal toughness in subclasses of weakly chordal graphs

The toughness of a graph $G$ is defined as the largest real number $t$ such that for any set $S\subseteq V(G)$ such that $G-S$ is disconnected, $S$ has at least $t$ times more elements than $G-S$ has components (unless $G$ is complete, in which case the toughness is defined to be infinite). A graph is said to be minimally tough if deleting any edge decreases the toughness. It is an open question whether there exists a minimally tough non-complete chordal graph with toughness exceeding $1$. We initiate the study of minimally tough graphs in the larger class of weakly chordal graphs. We obtain complete classifications of minimally tough graphs in the following subclasses of weakly chordal graphs: co-chordal graphs whose complement has diameter at least $3$, net-free co-chordal graphs, complements of forests, $P_4$-free graphs, and complete multipartite graphs. Our approach leads to simple proofs of two results on minimally tough graphs due to Dallard, Fernández, Katona, Milanič, and Varga.
Read the full abstract

Analysis of PDEs

Dispersion for the Schrödinger equation on the line with short-range array of delta potentials

We study dispersive properties of the one-dimensional Schrödinger equation with a short-range array of delta interactions. More precisely, we consider the self-adjoint operator obtained by perturbing the free Laplacian on the line with a real-valued sequence of Dirac delta potentials and belonging to weighted ${\ell}$^1(Z) spaces. Under suitable decay assumptions on the coupling constants and in the absence of a zero-energy resonance, we establish the L^1 (R) $\rightarrow$ L^$\infty$ (R) dispersive estimate with decay rate |t|^{-1/2} for the associated Schrödinger group. The proof relies on a limiting absorption principle in weighted spaces, explicit representation of the resolvent kernel in terms of Jost solutions and Born series expansion of the Friedrichs extension of the perturbed operator.
Read the full abstract

Probability

Quantitative Error Estimates for Learning Macroscopic Mobilities from Microscopic Fluctuations

We develop quantitative error estimates connecting microscopic fluctuation of interacting particle systems with the mobilities of their hydrodynamic limits. Focusing on the Symmetric Simple Exclusion Process and systems of independent Brownian particles, we provide explicit bounds for the discrepancy between the quadratic variation of fluctuation fields and the corresponding mobilities, in terms of time and spatial discretization parameters. In addition, we establish analogous error estimates for a class of fluctuating hydrodynamic stochastic PDEs with regularized coefficients. For stochastic PDEs with irregular square-root type coefficients, including Dean-Kawasaki type equations, we further identify the asymptotic behavior of the associated fluctuation structures within the framework of renormalized kinetic solutions. Our results provide quantitative insights into the relationship between microscopic fluctuation mechanisms and macroscopic mobilities, and contribute to a structured comparison between discrete particle systems and continuum fluctuating hydrodynamic descriptions.
Read the full abstract

Mathematical Physics · Cross-listed from quant-ph

Parsimonious Quantum Low-Density Parity-Check Code Surgery

Quantum code surgery offers a flexible, low-overhead framework for executing logical measurements within quantum error-correcting codes. It encompasses several fault-tolerant logical computation schemes, including parallel surgery, universal adapters and fast surgery, and serves as the key primitive in extractor architectures. The efficiency of these schemes crucially depends on constructing low-overhead ancilla systems for measuring arbitrary logical operators in general quantum Low-Density Parity-Check (qLDPC) codes. In this work, we introduce a method to construct an ancilla system of qubit size $O(W \log W)$ to measure an arbitrary logical Pauli operator of weight $W$ in any qLDPC stabilizer code. This new construction immediately reduces the asymptotic overhead across various quantum code surgery schemes.
Read the full abstract

All authors Andrew C. Yuan, Alexander Cowtan, Zhiyang He, Ting-Chun Lin, Dominic J. Williamson

Optimization and Control

Boundary stabilization of flows in networks of open channels modeled by Saint-Venant equations

This work investigates the boundary stabilization of flows in star-shaped and tree-shaped networks of open channels governed by the Saint-Venant equations with a friction term. Due to the existence of the friction term, the steady-states are non-uniform. We show that any such network can be stabilized with only controls at the terminal nodes of the network, even when there are no controls at the nodes inside the network. The number of control is optimal. The main tool we use is the Lyapunov approach, and the main challenge is that the state-of-the-art Lyapunov functions developed for Saint-Venant equations with source terms cannot be used. In this work, we manage to construct a new efficient and explicit Lyapunov function and, in turn, we give explicit ranges of the control tuning parameters that depend only on the values of the given non-uniform steady-states at the ends of the branches. Moreover, this Lyapunov function also improves the existing conditions found in the last decade for a single channel modelled by Saint-Venant equations.
Read the full abstract

Geometric Topology

Brunnian spanning 3-disks for the 2-unlink in the 4-sphere

We show that the $2$-component unlink in $S^4$ admits infinitely many isotopy classes of spanning $3$-disks that are Brunnian.

Information Theory · Cross-listed from cs.LG

Asymptotic Behavior of Multi–Task Learning: Implicit Regularization and Double Descent Effects

Multi–task learning seeks to improve the generalization error by leveraging the common information shared by multiple related tasks. One challenge in multi–task learning is identifying formulations capable of uncovering the common information shared between different but related tasks. This paper provides a precise asymptotic analysis of a popular multi–task formulation associated with misspecified perceptron learning models. The main contribution of this paper is to precisely determine the reasons behind the benefits gained from combining multiple related tasks. Specifically, we show that combining multiple tasks is asymptotically equivalent to a traditional formulation with additional regularization terms that help improve the generalization performance. Another contribution is to empirically study the impact of combining tasks on the generalization error. In particular, we empirically show that the combination of multiple tasks postpones the double descent phenomenon and can mitigate it asymptotically.
Read the full abstract

Logic · Cross-listed from cs.LO

Modal Fragments

We survey systematic approaches to basis-restricted fragments of propositional logic and modal logics, with an emphasis on how expressive power and computational complexity depend on the allowed operators. The propositional case is well-established and serves as a conceptual template: Post's lattice organizes fragments via Boolean clones and supports complexity classifications for standard reasoning tasks. For modal fragments, we then bring together two historically independent lines of investigation: a general framework where modal fragments are parameterized by a basis of "connectives" defined by arbitrary modal formulas (initially proposed and studied by logicians such as Kuznetsov and Ratsa in the 1970s), and the more tractable class of what we call simple modal fragments parameterized by Boolean functions plus selected modal operators, where Post-lattice methods enable systematic decidability and dichotomy results. Along the way, we collect and extend results on teachability and exact learnability from examples for both propositional fragments and simple modal fragments, and we conclude by identifying several open problems.
Read the full abstract

All authors Nick Bezhanishvili, Balder ten Cate, Arunavo Ganguly, Arne Meier

Probability

Regularization by noise for Gevrey well-posedeness of a weakly hyperbolic operator

We present an example of a linear partial differential equation whose Cauchy problem becomes well-posed when perturbed by noise. Specifically, we make clear how a suitable multiplicative Stratonovich perturbation of Brownian type renders a weakly hyperbolic operator with double involutive characteristics well-posed in the $C^{\infty}$-category, while its deterministic counterpart is only well-posed in the Gevrey $ s $ classes with $ 1 \leq s <2 $ .
Read the full abstract

Complex Variables

Bloch and Landau constants for meromorphic functions

Let $\mathcal{M}_1(\lambda)$ be the class of all meromorphic functions $f$ in the unit disk $\mathbb{D}=\{z\in\mathbb{C}\}: |z|<1$ having a simple pole at $\lambda \in \overline{\mathbb{D}} \setminus \{0\}$ and satisfying the normalization $f'(0)=1$. Let $B(\lambda)$ and $L(\lambda)$ denote the Bloch and Landau constants, respectively, for this class. In this article, we first show that the Bloch constant $B(1)$ and the Landau constant $L(1)$ are infinite. Using these results and a conformal mapping technique, we establish that $B(p)$ and $L(p)$ are likewise infinite for any $p \in (0,1)$, thereby refuting a recent conjecture. Finally, we extend our study to the class of meromorphic functions having two simple poles and prove that their associated Bloch and Landau constants also remain infinite.
Read the full abstract

Number Theory

A stabilizer interpretation of the (extended) linearized double shuffle Lie algebra

The linearized double shuffle Lie algebra introduced by Brown reflects the depth-graded structure of multiple zeta values. In a previous paper, the first author introduced an extension of this Lie algebra that accommodates multiple q-zeta values and multiple Eisenstein series. Inspired by the stabilizer interpretation of the double shuffle Lie algebra given by Enriquez and Furusho, we provide in this paper a stabilizer interpretation of both Lie algebras and show that the stabilizers preserve the extension from the first linearized Lie algebra to the second one.
Read the full abstract

Optimization and Control

Robust adaptive NMPC using ellipsoidal tubes

We propose a computationally efficient nonlinear Model Predictive Control (NMPC) algorithm for safe, learning-based control. The system model is represented as an affine combination of basis functions with unknown parameters, and is subject to additive set-bounded disturbances. Our algorithm employs successive linearization around nominal predicted trajectories and accounts for uncertainties in predicted states due to linearization, model errors, and disturbances using ellipsoidal sets. The ellipsoidal tube-based approach ensures that constraints on control inputs and system states are satisfied. Robustness to uncertainty is ensured using bounds on linearization errors and a backtracking line search. We show that the ellipsoidal embedding of model uncertainty scales favourably with system dimensions in numerical simulations. The algorithm incorporates set membership parameter estimation, and provides guarantees of recursive feasibility and input-to-state practical stability.
Read the full abstract

Optimization and Control

A Second-Order Algorithm Based on Affine Scaling Interior-Point Methods for nonlinear Optimisation with bound constraints

The homogeneous second-order descent method (Zhang et al. 2025, Mathematics of Operations Research) was initially proposed for unconstrained optimisation problems. HSODM shows excellent performance with respect to the global complexity rate among a certain broad class of second-order methods. In this paper, we extend HSODM to solve nonlinear optimisation problems with bound constraints and propose a second-order algorithm based on affine scaling interior-point methods (SOBASIP). In each iteration, an appropriate affine matrix is introduced to construct an affine scaling subproblem based on the optimality conditions of the problem. To obtain a valid descent direction similar to HSODM, we utilise the homogenisation technique to transform the scaling subproblem into an Ordinary Homogeneous Model (OHM), which is essentially an eigenvalue problem that can be solved efficiently. The descent direction is constructed from the optimal solution to the OHM, and then backtracking line search is used to determine the new iteration point. Theoretical analysis establishes that SOBASIP achieves a global iteration complexity of $O(\epsilon^{-3/2})$ for finding an $\epsilon$-approximate second-order stationary point and converges locally at a superlinear rate under certain conditions. Numerical results demonstrate that the proposed method exhibits satisfactory performance.
Read the full abstract

All authors Yonggang Pei, Yubing Lin, Mauricio Silva Louzeiro, Detong Zhu

Information Theory · Cross-listed from eess.SY

Formal Entropy-Regularized Control of Stochastic Systems

Analyzing and controlling system entropy is a powerful tool for regulating predictability of control systems. Applications benefiting from such approaches range from reinforcement learning and data security to human-robot collaboration. In continuous-state stochastic systems, accurate entropy analysis and control remains a challenge. In recent years, finite-state abstractions of continuous systems have enabled control synthesis with formal performance guarantees on objectives such as stage costs. However, these results do not extend to entropy-based performance measures. We solve this problem by first obtaining bounds on the entropy of system discretizations using traditional formal-abstractions results, and then obtaining an additional bound on the difference between the entropy of a continuous distribution and that of its discretization. The resulting theory enables formal entropy-aware controller synthesis that trades predictability against control performance while preserving formal guarantees for the original continuous system. More specifically, we focus on minimizing the linear combination of the KL divergence of the system trajectory distribution to uniform – our system entropy metric – and a generic cumulative cost. We note that the bound we derive on the difference between the KL divergence to uniform of a given continuous distribution and its discretization can also be relevant in more general information-theoretic contexts. A set of case studies illustrates the effectiveness of the method.
Read the full abstract

Mathematical Physics

Causal Fermion Systems, Non-Commutative Geometry and Generalized Trace Dynamics

We compare the structures and methods in the theory of causal fermion systems with generalized trace dynamics and non-commutative geometry. Although the three theories differ on many aspects, they agree in that the geometric structure to be recovered in the continuum limit is not the bare spacetime but a suitable fiber bundle. Furthermore, the comparison leads us to the conclusion that the key innovation in causal fermion systems lies in the manner in which the relation between different spacetime points is encoded. The role of Synge's classical world function $\sigma(x,y)$ that encodes the geodesic distance between any two points in the manifold, is taken by a generalized two-point correlator. We show that this idea can be transferred to non-commutative geometry and generalized trace dynamics.
Read the full abstract

All authors Shane Farnsworth, Felix Finster, Claudio F. Paganini, Tejinder P. Singh

Analysis of PDEs

Limiting absorption principle for time-harmonic acoustic and electromagnetic scattering of plane waves from a bi-periodic inhomogeneous layer

The Rayleigh expansion is widely used as a formal radiation condition in the analysis and numerical treatment of grating diffraction problems for incoming plane waves. However, the Rayleigh expansion does not always lead to uniqueness of open waveguide scattering problems, due to the existence of surface/guided waves (in other words, Bound States in the Continuum (BICs)) which exponentially decay in the direction perpendicular to the periodicity. In this paper we suppose that a bi-periodic inhomogeneous medium supports BICs at some real-valued incident wavenumber. Based on singular perturbation arguments, we justify the Limiting Absorption Principle (LAP) for both time-harmonic acoustic and electromagnetic scattering of plane waves from bi-periodic structures. Replacing the wavenumber $k$ with $k+i\epsilon$, we prove that the unique solution with $\epsilon>0$ converges to a solution of the original diffraction problem that additionally satisfies an orthogonal identity. This constraint condition together with the classical Rayleigh expansion leads to a sharp radiation condition to ensure uniqueness of time-harmonic scattering of plane waves by BIC-supporting bi-periodic materials.
Read the full abstract

Numerical Analysis

Nitsche methods for constrained problems in mechanics

We present guidelines for deriving new Nitsche Finite Element Methods to enforce equality and inequality constraints that act on the value of the unknown mechanical quantity. We first formulate the problem as a stabilized finite element method for the saddle point formulation where a Lagrange multiplier enforces the underlying constraint. The Nitsche method is then presented in a general minimization form, suitable for adding constraints to nonlinear finite element methods and allowing straightforward computational implementation with automatic differentation. This extends the method beyond classical boundary condition enforcement. To validate these ideas, we present Nitsche formulations for a range of problems in solid mechanics and give numerical evidence of the convergence rates of the Nitsche method.
Read the full abstract

All authors Tom Gustafsson, Antti Hannukainen, Vili Kohonen, Juha Videman

Combinatorics · Cross-listed from cs.DM

Positional s-of-k games

We introduce a general framework for positional games in which players score points by claiming a prescribed portion of each winning set, extending the notion of scoring Maker-Breaker games. In the scoring variant, Maker gains a point by fully claiming a winning set, while Breaker aims to minimize Maker's total score. In this paper, we generalize these models for all k-uniform positional games by fixing an integer threshold s in {1,2,..., k} so that a player scores a point whenever she claims at least s elements of a winning set of size k. We refer to this class as s-of-k games. Such formulation allows for a flexible description of scoring objectives that appear in both theoretical models and real-life board games. We further investigate the impact of strategy restrictions on the achievable score. In particular, we analyze s-of-k games both under optimal play, where the score is denoted by SC, and under the additional constraint that Maker is restricted to a pairing strategy. The corresponding score in this setting is denoted by SC_2. While the unrestricted score captures the standard notion of optimal play in scoring positional games, the pairing-restricted score allows us to observe Maker's loss incurred by limiting her to these standard strategies. We comprehensively study s-of-k games played on regular grids, which provide a natural and uniform setting for illustrating the general framework. After developing several general tools for the analysis of both scores, we complement them by a number of ad-hoc strategies tailored for particular cases of these games, to obtain both upper and lower bounds for the two scores on triangular, square, rhombus and hexagonal grids.
Read the full abstract

Optimization and Control · Cross-listed from cs.LG

Non-Euclidean Gradient Descent Operates at the Edge of Stability

The Edge of Stability (EoS) is a phenomenon where the sharpness (largest eigenvalue) of the Hessian approaches and then hovers near the stability threshold $2/\eta$ during gradient descent (GD) with step size $\eta$. Despite (apparently) violating classical smoothness assumptions, EoS has been widely observed in deep learning, but its theoretical foundations remain incomplete. We provide an interpretation of EoS through the lens of Directional Smoothness [Mishkin et al., 2024]. This interpretation naturally extends to non-Euclidean norms, which we use to define generalized sharpness under an arbitrary norm. Our generalized sharpness measure includes previously studied vanilla GD and preconditioned GD as special cases, as well as methods for which EoS has not been studied, such as $\ell_{\infty}$-descent, Block CD, Spectral GD, and their normalized versions. Through experiments on neural networks, we show that non-Euclidean GD with our generalized sharpness also exhibits progressive sharpening followed by oscillations around or above the threshold $2/\eta$. Practically, our framework provides a geometry-aware spectral diagnostic that can be applied across a broad class of non-Euclidean gradient methods.
Read the full abstract

All authors Rustem Islamov, Michael Crawshaw, Jeremy Cohen, Robert Gower

Optimization and Control

Stochastic Optimal Feedforward-Feedback Control for Partially Observable Sensorimotor Systems

Robust control of complex engineered and biological systems hinges on the integration of feedforward and feedback mechanisms. This is exemplified in neural motor control, where feedforward muscle co-contraction complements sensory-driven feedback corrections to ensure stable behaviors. However, deriving a general continuous-time framework to determine such optimal control policies for partially observable, stochastic, nonlinear, and high-dimensional systems remains a formidable computational challenge. Here, we introduce a framework that extends neighboring optimal control by enabling the feedforward plan to explicitly account for feedback uncertainties and latencies. Using statistical linearization, we transform the stochastic problem into an approximately equivalent deterministic optimization within a tractable, augmented state space that retains critical nonlinearities, offering both mechanistic interpretability and theoretical guarantees on approximation fidelity. We apply this framework to human neuromechanics, demonstrating that muscle co-contraction emerges as an optimal adaptation to task demands, given the characteristics of our sensorimotor system. Our results provide a computational foundation for neuromotor control and a generalizable tool for the control of nonlinear stochastic systems.
Read the full abstract

Information Theory

On LLR Mismatch in Belief Propagation Decoding of Overcomplete QLDPC Codes

Belief propagation (BP) decoding of quantum low density parity check (QLDPC) codes is often implemented using overcomplete stabilizer (OS) representations, where redundant parity checks are introduced to improve finite length performance. Decoder behavior for such representations is governed primarily by finite iteration dynamics rather than asymptotic code properties. These dynamics are known to critically depend on the initialization of the decoder. In this paper, we investigate the impact of mismatched log likelihood ratios (LLRs) used for BP initialization on the performance of QLDPC codes with OS representations. Our results demonstrate that initial LLR mismatch has a strong influence on the frame error rate (FER), particularly in the low noise regime. We also show that the optimal performance is not sharply localized: the FER remains largely insensitive over an extended region of mismatched LLRs. This behavior motivates an interpretation of LLR mismatch as a regularization control parameter rather than a quantity that must be precisely matched to the quantum channel.
Read the full abstract

All authors Hernan Cordova, Alexios Balatsoukas-Stimming, Gabriele Liga, Yunus Can Gültekin, Alex Alvarado

Analysis of PDEs

$\mathrm{L}^{2}$–convergence of the time-splitting scheme for nonlinear Dirac equation in 1+1 dimensions

We study the time-splitting scheme for approximating solutions to the Cauchy problem of the nonlinear Dirac equation in 1+1 dimensions. Under the assumption that the initial data for the scheme are convergent in $\mathrm{L}^{2}(\mathbb{R})$, we prove that the approximate solutions constructed by the corresponding time-splitting scheme are strongly convergent in $\mathrm{L}^{2}(\mathbb{R}\times[0,T])$ to the global strong solution of the nonlinear Dirac equation for any $T>0$. To achieve this, we first establish the pointwise estimates for time-splitting solutions. Based on these estimates, a modified Glimm-type functional is carefully designed to show that it is uniformly bounded in time, which yields $\mathrm{L}^2$ stability estimates for the scheme. Furthermore, we prove that the set of time-splitting solutions is relatively compact in $\mathrm{C}([0,T];\mathrm{L}^{2}(\mathbb{R}))$ for any $T>0$. Finally, we show that the limit of any convergent subsequence of the time-splitting solutions is the strong solution to the Cauchy problem of the nonlinear Dirac equation.
Read the full abstract

Algebraic Geometry

Lagrangian structures on the derived moduli of constructible sheaves

Given a compact oriented manifold of dimension $n$ with a conically smooth stratification, we show that the moduli of $\mathcal{D}(k)$-valued constructible sheaves and the moduli of perverse sheaves are $(2-n)$-shifted Lagrangian. The former statement follows from the construction of a relative left $n$-Calabi–Yau structure on the stable $\infty$-category of $\mathcal{D}(k)$-valued constructible sheaves. This is achieved via a lax gluing result for categorical cubes equipped with cubical Calabi–Yau structures. Given a codimension $2$ submanifold, we further identify symplectic leaves corresponding to perverse sheaves with prescribed monodromy.
Read the full abstract

Complex Variables

Analytic structure of $q$-pseudoconcave subsets of continuous graphs

The aim of this article is to find under which conditions there exists a foliation by $n$-dimensional complex manifolds on a closed subset $Z\subset\mathbb{C}^N$, which is locally a continuous graph over a closed subset of $\mathbb{C}^n\times\mathbb{R}$. We prove that such foliation does exist if $Z$ is $q$-pseudoconcave, for any $q\geq n$. We also prove that this bound on the index of pseudoconcavity is sharp. Namely, we construct many $q$-pseudoconcave subsets of $\mathbb{C}^N$, for $q<n$, which are smooth graphs over closed subsets of $\mathbb{C}^n\times\mathbb{R}$ but with no analytic structure. As an application, we show that an $n$-pseudoconcave set sitting in a subset of $\mathbb{C}^N$, which is locally a differentiable graph over a domain of $\mathbb{R}^{2n+1}$, is foliated by $n$-dimensional complex manifolds.
Read the full abstract

Combinatorics

Quadratic form estimations for Hessian matrices of resistance distance and Kirchhoff index of positive-weighted graphs

Let $G^{w}=(V,E,w)$ be a positive-weighted graph with the weight $w(e)>0$ for all $e\in E$. The weighted graph $G^{\widetilde{w}}=(V,E,\widetilde{w})$ is called a hyper-dual number weighted graph, where the weight $\widetilde{w}(e)=w(e)+\Delta w(e)(\varepsilon+\varepsilon^{*})$ is a hyper dual number, $\Delta w(e)$ is a real number, $\varepsilon$ and $\varepsilon^{*}$ are two dual units, $e\in E$. In this paper, we give a representation for the Moore-Penrose inverse of the Laplacian matrix, and calculation formulas for the resistance distance and Kirchhoff index of $G^{\widetilde{w}}$, respectively. We establish quadratic forms of the Hessian matrices for the resistance distance and Kirchhoff index of $G^{w}$ via generalized matrix inverses. We further derive explicit bounds on the eigenvalues of the Hessian matrices for the resistance distance and the Kirchhoff index of $G^{w}$ in terms of graph parameters. We also prove that the Kirchhoff index of a positive-weighted graph with bounded edge weights is strongly convex on its edge weight vector.
Read the full abstract

Information Theory · Cross-listed from cs.LG

WaterSIC: Information-Theoretically (Near) Optimal Linear Layer Quantization

This paper considers the problem of converting a given dense linear layer to low precision. The tradeoff between compressed length and output discrepancy is analyzed information theoretically (IT). It is shown that a popular GPTQ algorithm may have an arbitrarily large gap to the IT limit. To alleviate this problem, a novel algorithm, termed “WaterSIC”, is proposed and is shown to be within a rate gap of 0.255 bits to the IT limit, uniformly over all possible covariance matrices of input activations. The key innovation of WaterSIC's is to allocate different quantization rates to different columns (in-features) of the weight matrix, mimicking the classical IT solution known as "waterfilling". Applying WaterSIC to the Llama and Qwen family of LLMs establishes new state-of-the-art performance for all quantization rates from 1 to 4 bits. Our code is available at https://github.com/egorlifar/watersic.
Read the full abstract

All authors Egor Lifar, Semyon Savkin, Or Ordentlich, Yury Polyanskiy

Optimization and Control

An Efficient Stochastic First-Order Algorithm for Nonconvex–Strongly Concave Minimax Optimization beyond Lipschitz Smoothness

In recent years, nonconvex minimax problems have attracted significant attention because of their broad applications in machine learning, including generative adversarial networks, robust optimization and adversarial training. Most existing algorithms for nonconvex stochastic minimax problems are developed under the standard Lipschitz smoothness assumption. In this paper, we study stochastic minimax problems under a generalized smoothness condition and propose an algorithm, NSGDA-M, which simultaneously updates the inner variable by stochastic gradient ascent and updates the outer variable by normalized stochastic gradient descent with momentum. When the objective function is nonconvex–strongly concave, we show that NSGDA-M finds an \(\epsilon\)-stationary point of the primal function within \(\mathcal O(\epsilon^{-4}\log(1/(\epsilon\delta)))\) stochastic gradient evaluations with probability at least \(1-\delta\). Moreover, we establish an expected stationarity guarantee of \(\mathcal O(\delta^{-3/4}T^{-1/4})+G_\Phi\delta\), which gives a convergence rate \(\mathcal O(T^{-1/7})\). Here \(G_\Phi\) bounds the primal gradient norms along the iterates. Numerical experiments on a distributionally robust optimization problem demonstrate the effectiveness of the proposed algorithm.
Read the full abstract