Information Theory
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
Tuesday 12 May
Solomon and Stiffler constructed infinitely many families of linear codes meeting the Griesmer bound in 1965. It is well-known in 1990's that certain Griesmer codes (codes with the zero Griesmer defect) are equivalent to Solomon-Stiffler codes or Belov codes. Griesmer codes constructed in some recent papers published in IEEE Trans. Inf. Theory are actually Solomon-Stiffler codes or affine Solomon-Stiffler codes proposed in our previous paper. Therefore it is more challenging to construct optimal codes with positive Griesmer defects. In this paper, we construct several infinite families of optimal codes with positive Griesmer defects. Then these codes are certainly not equivalent to Solomon-Stiffler codes or Belov codes. Weight distributions and subcode support weight distributions of these optimal codes are determined. On the other hand, some of constructed optimal linear codes are optimal locally recoverable codes (LRCs) meeting the Cadambe-Mazumdar (CM) bound. Some of our constructed optimal codes are very close to the CM bound. Localities of these optimal or almost optimal LRC codes are two.
Optimization and Control
From Schrodinger Bridge to Optimal Transport over Sub-Riemannian Manifolds
We study the least-energy way to reshape a probability distribution when motion is constrained to a horizontal bundle, that is, optimal transport and distribution steering in sub-Riemannian geometry, motivated by density control over underactuated systems. To obtain a continuous and numerically tractable formulation, we introduce an entropic regularization by adding small noise aligned with the control directions and study the associated Schrodinger bridge problem. The resulting reference process is a degenerate diffusion on the sub-Riemannian manifold. Under bracket-generating hypotheses we obtain smooth, strictly positive transition densities and a forward–backward characterization of the optimal bridge. This leads to a practical Sinkhorn-type algorithm for the Schrodinger potentials and, as the noise level vanishes, a recovery of the deterministic sub-Riemannian optimal transport problem. We demonstrate with a numerical example.
Read the full abstract
Optimization and Control
The Grimmer–Shu–Wang Certificate and the Drori–Teboulle Minimax Constant-Stepsize Bound for $N\ge 3$
We prove, for every horizon \(N\ge 3\), the existence of the strengthened low-rank performance-estimation certificate proposed by Grimmer, Shu, and Wang for the Drori–Teboulle constant-step gradient-descent bound. For each \(N\ge 3\), let \(\rho_N\in(0,1)\) be determined by \(\rho_N^{2N}(2N\rho_N+2N+1)=1\). We show that the GSW certificate equations admit positive vectors \(a,b,c,d\) satisfying all residual equations. The proof proceeds through a reduced residual system in the variables \(d\), a simplex existence argument for a positive reduced zero, a terminal residual completion identity, and a tail-square convolution argument proving the cumulative margins that force \(b>0\) and then \(a>0\). Consequently, the GSW low-rank PEP certificate exists for every \(N\ge 3\) and yields the Drori–Teboulle upper bound. We also include the one-dimensional quadratic and Huber lower-bound examples for nonnegative steps, while the quadratic example excludes negative constant steps from being optimal. Together these prove the Drori–Teboulle minimax constant-stepsize value over all real constant steps for every \(N\ge 3\).
Read the full abstract
Mathematical Physics · Cross-listed from gr-qc
Dynamics of a relativistic discrete body: rigidity conditions, and covariant equations of motion
Rigidity conditions for a body considered as a discrete system of relativistic particles are proposed. They by themselves do not yet determine an evolution of the system, and some second-order equations must be added to them. Poincaré-covariant equations of motion compatible with these rigidity conditions are proposed and discussed. The resulting theory has the expected six dynamical degrees of freedom, allowing for more general motions than in Born's theory. Therefore, treating a relativistic body as a discrete system of particles could be a promising alternative to the standard approach based on Born's rigidity conditions.
Read the full abstract
Numerical Analysis
Inverse initial data for nonlinear Schrödinger equation via Carleman estimates and the contraction principle
We study an inverse initial-data problem for a nonlinear Schrödinger equation in which the initial wave field is reconstructed from lateral measurements. Our approach combines a Legendre-polynomial-exponential-time dimensional reduction with a Carleman-based contraction principle. First, we expand the solution in a weighted Legendre basis in time and truncate the expansion to obtain a coupled nonlinear elliptic system for the spatial coefficients. Next, we solve this reduced system by constructing a contraction map on a suitable admissible set. This contraction map admits a unique fixed point, which is the limit of the corresponding Picard iteration. We also establish a stability estimate showing that this fixed point remains close to the exact reduced solution in the noisy-data case. Finally, we present numerical experiments in two space dimensions for several different geometries and nonlinear exponents. The numerical results show that the proposed method accurately reconstructs the main features of the initial wave field and remains stable even when the boundary data contain noise.
Read the full abstract
Combinatorics
On minimal collections of sequences for testing continuity
We study test sets: subfamilies of sequences converging to a point P that still suffice to detect every discontinuity of real-valued functions at P. Ordered by inclusion, these test sets form a poset. Under natural hypotheses at P, we prove that this poset has a minimal element. We also analyze its maximal chains, showing that some have a least element, while others do not. Finally, on the sequential fan we give a concrete realization in which the minimal test set produced by our construction has strictly smaller cardinality than the full family of convergent sequences.
Read the full abstract
Algebraic Topology
Obstructions for Associativity in Stable Homotopy Theory
We give a construction of the obstruction theory for $\mathbb{A}_{n}$-algebra structures in stable $\infty$-categories, and give some properties of it. We use this to show that the spectrum $\mathbb{S} / 4$ admits an $\mathbb{A}_5$-multiplication using synthetic spectra.
Dynamical Systems
Bistability, Absolute Concentration Robustness, and Hysteresis in Dual-Site Futile Cycles with Bifunctional Enzymes
Bifunctional enzymes, which catalyze both the forward and reverse steps of a substrate modification reaction, arise naturally in bacterial two-component signaling systems and metabolic regulation. Beyond their well-known role in conferring absolute concentration robustness (ACR) on substrate species, bifunctional enzymes profoundly shape the dynamical landscape of the networks in which they appear. We study a class of dual-site futile cycles in which the reverse modification steps are carried out by bifunctional enzyme-substrate compounds, and provide a complete mathematical analysis of all four such networks, characterizing the existence, number, and stability of steady states, as well as the bifurcation structure as total substrate is varied. All four networks admit boundary steady states, in contrast to the non-bifunctional case. The networks differ in the number and stability of boundary steady states, in the maximum number of positive steady states (ranging from two to four), and in whether bistability is present. In two networks, a transcritical bifurcation connects the boundary and positive steady state branches; in one case this is a backward bifurcation, producing hysteresis. Perhaps the most striking phenomenon occurs in one of the four networks, which simultaneously exhibits bistability and ACR in the final modification state, where the system can settle into either of two stable steady states with different intermediate concentrations yet identical final product concentration.
Read the full abstract
Differential Geometry
Rigidity and flexibility under spectral Ricci lower bounds and mean-convex boundary
We study Riemannian manifolds $(M^n,g)$ with mean-convex boundary whose Ricci curvature is nonnegative in a spectral sense. Our first main result is a sharp spectral extension of a rigidity theorem by Kasue: we prove that under the conditions \[ \lambda_1(-\gamma\Delta+\mathrm{Ric})\geq 0,\qquad H_{\partial M}\geq 0, \] and in the sharp range $0\leq \gamma<4$ if $n=2$, and $0\leq\gamma<\frac{n-1}{n-2}$ if $n\geq3$, a (possibly noncompact) complete manifold with disconnected boundary, with at least one compact boundary component, must split isometrically as a product $[0,L]\times \Sigma$. Our second main contribution is a topological rigidity result for the relative fundamental group $\pi_1(M,\partial M)$, combined with a deep theorem of Lawson–Michelsohn. We prove that, in dimensions $n\neq4$, any compact manifold with boundary satisfying the two inequalities above, with at least one of them strict, admits a metric with positive sectional curvature and strictly mean-convex boundary, provided $\gamma\geq0$ if $n=2$, and $0\leq\gamma\leq\frac{n-1}{n-2}$ if $n\geq3$. This range of $\gamma$ is sharp for the latter result to hold.
Read the full abstract
Statistics Theory
The Geometry of Spectral Fluctuations: On Near-Optimal Conditions for Universal Gaussian CLTs, with Statistical Applications
We study linear spectral statistics of high dimensional sample covariance matrices in a regime where the empirical spectral distribution remains governed by the classical sample covariance law but the fluctuation theory is nonclassical. Our starting point is a decomposition of the covariance of centered quadratic forms into a universal Gaussian part and a model dependent fourth order correction. This leads to an abstract framework, termed GHOST, for universal Gaussian central limit theorems under structured fourth order effects. Under this framework, we prove a Gaussian central limit theorem for linear spectral statistics, with explicit mean and covariance corrections determined by a bilinear fourth order kernel. Boundary examples show that the conditions are close to necessary for a broad universal Gaussian closure. We then develop a blockwise mixed radial model that verifies the abstract assumptions and makes the correction explicit. The correction splits into an entrywise fourth moment component and a lockwise energy fluctuation component. The latter may change the fluctuation scale, leading to a phase transition at the level of fluctuations. As an application, we study sphericity testing. Under the spherical null, the general correction collapses to a single scalar parameter, yielding a feasible data driven correction of John's test.
Read the full abstract
Optimization and Control
Reputation-Based Byzantine-Resilient Vector Consensus without Prespecified Fault Bounds
This paper proposes an adaptive reputation-based algorithm for Byzantine-resilient vector consensus in distributed multi-agent systems. The proposed algorithm does not require a prespecified bound on the number of Byzantine neighbors as an algorithmic input. Instead, each agent evaluates neighboring behavior from observed state discrepancies and updates reputation weights using accumulated behavioral evidence. The reputation mechanism adaptively redistributes communication weights and can assign exactly zero weight to sufficiently inconsistent neighbors. These reputation weights are incorporated into the local consensus updates to suppress adversarial influence while preserving reliable neighbor interactions. The resulting algorithm couples reputation adaptation with consensus dynamics while supporting high-dimensional vector states with low per-agent computational complexity. Extensive experiments on distributed systems against classical resilient consensus methods demonstrate improved Byzantine detection accuracy and more reliable consensus performance across different attack scenarios and state dimensions.
Read the full abstract
Information Theory
RankGuardPolar Private Public Finite Length Polar Codes with Rank-Certified Leakage
We introduce RankGuard-Polar, a framework for safely publishing a subset of polar codeword coordinates over shared public resources. We assume a strong eavesdropper who has access to the channel input, i.e., the transmitted codeword coordinates published on a public resource access model. Working over \(\mathbb F_2\) and focusing on time-shared public/private BEC uses, we show that leakage from a published index set \(\mathbf{P}\) admits an exact algebraic characterization comes from an information-theoretic viewpoint, and we construct an explicit linear extractor ($R$) that identifies the leaked linear combinations. Building on this identity, we (i) give efficient procedures to compute and certify leakage for any \(\mathbf{P}\), (ii) propose a practical fast algorithm with provable efficiency.
Read the full abstract
Information Theory
Parameter Estimation of Mutual Information Maximized Channels
We study the problem of estimating a parametric discrete memoryless channel \( p(y \mid x; \boldsymbol{\theta}) \) when the transmitter selects its input distribution \( \pi \) to maximize mutual information under the true parameter \( \boldsymbol{\theta}^* \). Using only i.i.d.\ observations of the channel output, we aim to jointly estimate the capacity-achieving input distribution \( \boldsymbol{\pi}^* \) and the true channel parameter \( \boldsymbol{\theta}^* \). In general, recovery of \( \boldsymbol{\pi}^* \) and \( \boldsymbol{\theta}^* \) can be challenging. To that end, we propose two efficient algorithms based on the Blahut–Arimoto (BA) optimality conditions: (i) a bilevel fixed-point method and (ii) an augmented Lagrangian method. Empirical results demonstrate that both proposed algorithms successfully recover the true \( \boldsymbol{\theta}^* \) and \( \boldsymbol{\pi}^* \), whereas a naive maximum-likelihood approach that ignores the mutual-information maximization constraint fails to do so.
Read the full abstract
Combinatorics
A short proof of Mathar's 2020 recurrence conjecture for the generalized-Stirling sequence A001711
For the OEIS sequence A001711, contributed by N. J. A. Sloane long before the on-line era and identified there as the diagonal $T(n+4, 4)$ of a generalized-Stirling triangle, R. J. Mathar contributed in February 2020 the conjectured order-2 P-recursive recurrence \[ a(n) - (2n+5)\,a(n-1) + (n+2)^{2}\,a(n-2) \;=\; 0,\qquad n \ge 2. \] We give a one-page proof. Detlefs's harmonic-number closed form $a(n) = \tfrac{1}{4}(n+3)!\,(2 H_{n+3} - 3)$ collapses the left-hand side, after dividing through by $(n+1)!/4$, to a polynomial identity of $n$ with coefficient $H_{n+2}$. The harmonic-number coefficient simplifies to $(n+3) - (2n+5) + (n+2) = 0$ (using $H_{n+3} = H_{n+2} + \tfrac{1}{n+3}$ and $H_{n+1} = H_{n+2} - \tfrac{1}{n+2}$); the constant remainder is $-3 \cdot 0 = 0$ for the same reason. The supplementary archive contains a SymPy script verifying both pieces symbolically, the e.g.f.\ expansion against the harmonic closed form, and Mathar's recurrence numerically for $n = 2, \ldots, 5000$.
Read the full abstract
Monday 11 May
Algebraic Geometry
Examples of descendent generating series for Pandharipande–Thomas stable pairs on smooth projective Fano threefolds via one-dimensional wall-crossing
We study descendent generating series for Pandharipande–Thomas stable pairs on smooth projective Fano threefolds. We use the wall-crossing setup developed by the author and Joyce in Joyce's Lie algebra $H_*(\N^{\pl},\Q)$ of the projective-linear pairs stack, and next pass to Gross's polynomial realization $e^{\kappa}\Q[s_{jk\ell}]$. We compute explicit examples of one-dimensional Donaldson–Thomas invariants on Fano 3-folds and, via wall-crossing, Pandharipande–Thomas stable pair invariants and descendent generating series. We compute examples on $\PP^3$, on a smooth cubic threefold, on $\Bl_p\PP^3$, on $\Bl_\ell\PP^3$, and on the projective-bundle threefold $\PP(\OO_X\oplus \OO_X(-1,-1))$ over $X=\PP^1\times\PP^1$. In the $\PP^3$ and cubic threefold examples we compare the intrinsic large-$n$ tails with the formulas of Pandharipande and Moreira and show that, in the cases treated in common, the differences are Laurent polynomials.
Read the full abstract
Mathematical Physics
A Metric-Deformed $q$-Gauge Dirac Equation
We construct a family of metric-deformed gauge theories based on a recently introduced $q$-Dirac operator $D_q = \gamma^\mu \sqrt{|g^{\mu\mu}|}\partial_\mu$, which arises from a deformed D'Alembertian $\Box_q = |g^{00}|\partial_t^2 - \sum_i |g^{ii}|\partial_i^2$. The deformation parameter $q$ is related to the metric components via $q_\mu = \sqrt{|g^{\mu\mu}|}$. By promoting $g^{\mu\mu}(x)$ to spacetime-dependent background fields, we define a deformed covariant derivative $D_\mu^{(q)} = \partial_\mu + ieA_\mu(x)/\sqrt{|g^{\mu\mu}(x)|}$ (no sum over $\mu$). The corresponding field strength $F_{\mu\nu}^{(q)} = [D_\mu^{(q)}, D_\nu^{(q)}]$ acquires new terms proportional to $\partial_\mu(1/\sqrt{|g^{\nu\nu}|})$, which vanish for constant metrics. We write down gauge-invariant actions for deformed Yang-Mills theory and for fermions minimally coupled to $D_\mu^{(q)}$. This work provides a mathematical foundation for $q$-deformed gauge theories from a metric perspective.
Read the full abstract
Information Theory · Cross-listed from cs.LG
Accurate Evaluation of Quickest Changepoint Detectors via Non-parametric Survival Analysis
We propose non-parametric estimators for the average run length (ARL) and average detection delay (ADD) in quickest changepoint detection (QCD) under finite and irregular sequence lengths. Although ARL and ADD are widely used as optimality criteria in theoretical and simulation studies, their application to real-world datasets is hindered by limited and irregular sequence lengths. To address this issue, we propose non-parametric estimators for the ARL and ADD, termed KM-ARL and KM-ADD, by drawing an analogy between QCD and survival analysis to model detection probabilities under sequence truncation. We derive estimation bias bounds and prove that they are asymptotically unbiased unless extrapolation is required. Experiments on simulated and real-world datasets demonstrate their practical utility, enhancing robustness against limited and irregular sequence lengths, improving interpretability, and facilitating empirical, intuitive model selection. Our Python code is provided at https://github.com/TaikiMiyagawa/Kaplan-Meier-Average-Run-Length, offering ready-to-use implementations for practitioners.
Read the full abstract
Analysis of PDEs
Hydrodynamic limit and Newtonian limit from the relativistic Vlasov-Maxwell-Boltzmann system to the classical Euler-Poisson system
In this paper, around a global smooth irrotational solution to the classical isentropic compressible Euler-Poisson system, we construct classical solutions to the one-species relativistic Vlasov-Maxwell-Boltzmann system on any finite time interval $[0,T]$, and rigorously justify the combined hydrodynamic and Newtonian limits to the Euler-Poisson system. In particular, this yields a rigorous derivation of the compressible Euler-Poisson system, whose Poisson coupling induces an instantaneous electrostatic response and thus no longer preserves a strict finite-speed propagation structure, from a relativistic kinetic model with finite propagation speed. The analysis is based on a Hilbert expansion in $\varepsilon$ for the relativistic Vlasov-Maxwell-Boltzmann system, an asymptotic expansion in $\mathfrak{c}^{-1}$ for the relativistic Euler-Maxwell system, and estimates for both the expansion coefficients and the remainder terms that are uniform in $\mathfrak{c}$ and $\varepsilon$, without imposing any a priori relation between these two parameters.
Read the full abstract
Information Theory · Cross-listed from cs.LG
An Information-Theoretic Criterion for Efficient Data Synthesis
Synthetic data becomes crucial for large language model training, but its effectiveness is highly inconsistent. We provide an information-theoretic account of this inconsistency: synthetic data improves a model only when the generation-training loop is information-open, i.e., shaped by external signals (verifiers, environments, or rubrics) that inject task-relevant information beyond the model's current distribution. When the loop is information-closed (relying on the model's own outputs without such signals), the data processing inequality ensures that task-relevant information can only decrease, making collapse a predicted outcome. Among information-open pipelines, both efficiency and generalization hinge on the meta-level of supervision: a coarser signal such as binary correctness treats all acceptable outputs as equivalent, so the behavior it teaches is not tied to any particular domain or surface form and generalizes naturally across tasks and domains. These observations lead to a guiding thesis: learning preferentially converges to the most information-efficient signal component available, which accelerates learning when that component is the intended one, but causes reward hacking when a spurious pattern happens to be simpler.
Read the full abstract
Number Theory
Parity vectors and paradoxical sequences in the accelerated Collatz map
This note studies parity vectors and paradoxical sequences in the accelerated Collatz iteration $T(n) = (3n+1)/2$ for $n$ odd, $T(n) = n/2$ for $n$ even. Building on Rozier and Terracol (arXiv:2502.00948, 2025), Terras (1976), Lagarias (1985), and Tao (2019), we prove three theorems and add one numerical observation. The first is a sharp finitary form of Terras's parity-vector density; the second is a closed-form analytic count of paradoxical $\Omega_k(n)$ for each fixed length $k$. The third is a density-zero theorem for bounded-length paradoxical sequences with explicit constant. As for the numerical piece, among the seven $(j, q)$ pairs that show up in the Rozier-Terracol enumeration with first term $n \le 10^9$, every paradoxical reduced ratio $q/j$ turns out to be a left convergent, a left semiconvergent, or a Stern-Brocot mediant of adjacent convergents/semiconvergents of $\log_3 2$. The three theorems are unconditional. The fourth observation is verified for $n \le 10^7$ and conjectured for all $n$. We make no claim toward the Collatz conjecture or Terras's coefficient-stopping-time conjecture.
Read the full abstract
Number Theory
A Generalized Closed Form of Ramanujan-Type Fourier Cosine Transform via Meijer's G-Function
In this paper, we obtain analytical evaluations of the Ramanujan integral \[\textbf{R}_{C}(m,n)= \int_{0}^{\infty}\frac{x^m\,\cos(\pi nx)}{\exp{(2\pi\sqrt{x})-1}}dx\] subject to suitable convergence conditions in terms of an infinite series of Meijer $G$-functions of one variable, by using Mellin-Barnes-type contour integral representations of the cosine function. %and Laplace transform method. We also consider some generalizations of the integral $\textbf{R}_{C}(m,n)$ given as the integrals $I_{C}^{*}(\upsilon,b,c,\lambda,y)$ ,$\Xi_{C}(\upsilon,b,c,\lambda,y)$, $\nabla_{C}(\upsilon,b,c,\lambda,y)$ and $I_{C}(\upsilon,b,\lambda,y)$. These integrals are also expressed in terms of infinite series of Meijer $G$-functions. Moreover, as an application of a Ramanujan's integral $\textbf{R}_C(m,n)$, the closed-form evaluations of nine infinite series of Meijer $G$-functions are obtained.
Read the full abstract
Algebraic Geometry
On the DCC Property of Iitaka Volume with Real Coefficients and Generalised Pairs
We investigate the DCC property of the set of Iitaka volumes of a given set of pairs of varieties. We both generalize previous results of Birkar and Li about usual pairs to the real coefficient case, and also establish similar results on generalised pairs, where some natural boundedness assumptions are required for technical reasons.
Read the full abstract
Commutative Algebra
Regularity under Integral Closure: A General Criterion and a Sharp Counterexample
We study the behavior of Castelnuovo–Mumford regularity under integral closure. We prove a general cohomological criterion for the inequality $\operatorname{reg}(\overline J)\leq \operatorname{reg(J)}$, and in particular obtain it for every homogeneous ideal in a polynomial ring in at most three variables. The proof uses a high-degree comparison between $\overline J$ and its saturation. We also give a saturated monomial counterexample in four variables, showing that the three-variable result is sharp.
Read the full abstract
Logic
Almost Disjointness Principles and $Q$-Space Cardinals
Banakh and Bazylevych introduced separation-axiom variants $\mathfrak q_i$, for $i=1,2,2\frac{1}{2}$, of the cardinal $\mathfrak q$, together with a cardinal $\mathfrak{adp}$ lying between $\mathfrak{dp}$ and $\mathfrak{ap}$. They asked whether $\mathfrak{adp}$ coincides with either of these two cardinals. We prove in ZFC that $\mathfrak{adp}=\mathfrak{dp}$. We define a dual variant $\mathfrak{adp}_2$ and show that $\mathfrak{adp}_2=\mathfrak{ap}$. We further study the relation between $\mathfrak{ap}$ and the weakened $Q$-space cardinals. We introduce a tree analogue $\mathfrak{at}$ of $\mathfrak{ap}$ and prove $\mathfrak{at}=\mathfrak q_{2}$, concluding that $\mathfrak{ap}\leq\mathfrak q_{2}$, which answers a question of Banakh and Bazylevych. Assuming the Generalized Continuum Hypothesis, we construct ccc forcing extensions with $\mathfrak{ap}=\omega_1<\mathfrak{at}=\mathfrak c$, so $\mathfrak{ap}<\mathfrak q_2$ is consistent with ZFC.
Read the full abstract
Commutative Algebra
Generalized Andrásfai graphs and special Betti diagrams of edge ideals
Edge ideals of graphs were introduced by Villarreal in 1990, and have been the subject of many studies since then. In the same year, Fröberg characterized edge ideals with regularity 2 in combinatorial terms. This result was generalized by Fernández-Ramos and Gimenez to regularity 3 for bipartite graphs. A key ingredient in these results is the particular shape of the Betti diagrams of the edge ideals of the graphs obtained after removing a Hamiltonian cycle from either a complete graph $ K_k$ or a complete bipartite graph $K_{k,k}$. In this work, we consider the family of Generalized Andrásfai graphs ${\rm GA}(t,k)$ with $t\geq 1 $ and $k \geq 2$. This family extends the families of complete graphs, since $K_{k+1} = {\rm GA}(1,k)$, and complete bipartite $k$-regular graphs, since $K_{k,k} = {\rm GA}(2,k)$. We show that the results known for $ K_k$ and $ K_{k,k}$ can be naturally extended to this family. More precisely, when removing a suitable Hamiltonian cycle from ${\rm GA}(t,k)$, the resulting edge ideal has regularity $t+2$, projective dimension $t(k-2)$ and a Betti diagram exhibiting a generalized version of the same special shape.
Read the full abstract
Probability
Positive probability of explosion for stochastic heat equation with superlinear accretive reaction term and polynomially growing multiplicative noise
This paper studies the finite time explosion of the stochastic heat equation $\frac{\partial u}{\partial t}(t,x)=\frac{\partial^2}{\partial x^2} u(t,x)+(u(t,x))^{\beta}+\sigma(u(t,x))\dot{W}(t,x)$. We consider an interval $D=[-\pi,\pi]$ under periodic boundary condition where $\dot{W}(t,x)$ is a space-time white noise and $\sigma(u)\approx u^{\gamma}$ near $\infty$. Our results refine existing results by identifying behavior in a previously less understood regime, where we show that if $\beta\in(1,3),\gamma\in(\frac{\beta}{2},\frac{\beta+3}{4})$ or $\beta>1,\gamma\in(0,\frac{\beta}{2}]$ then mild solutions can explode with positive probability. This paper provides a partial characterization of the explosion behavior in an intermediate parameter regime, and contribute to the understanding of the interplay between the drift and diffusion terms.
Read the full abstract
Optimization and Control · Cross-listed from cs.LG
Error whitening: Why Gauss-Newton outperforms Newton
The Gauss-Newton matrix is widely viewed as a positive semidefinite approximation of the Hessian, yet mounting empirical evidence shows that Gauss-Newton descent outperforms Newton's method. We adopt a function space perspective to analyze this phenomenon. We show that the generalized Gauss-Newton (GGN) matrix projects the Newton direction in function space onto the model's tangent space, while a Jacobian-only variant obtained by applying the least squares Gauss-Newton matrix to non-least squares losses projects the function space loss gradient onto this same tangent space. Both projections eliminate distortions from the model's parameterization. Specifically, the evolution of the prediction-target mismatch depends on the model's parameterization through the matrix $JJ^\top$ where $J$ is the Jacobian of the model with respect to its parameters. The projections effectively replace $JJ^\top$ with the identity. We call this effect error whitening. Once the parameterization is removed, the prediction-target mismatch evolves according to dynamics dictated by the structure of the loss and the projection produced by the optimizer. Error whitening is a special property of Gauss-Newton descent that rigorously distinguishes it from Newton's method. We empirically demonstrate that Gauss-Newton optimizers follow the theoretically predicted function space dynamics and outperforms Newton's method, Adam, and Muon across case studies spanning supervised learning, physics-informed deep learning, and approximate dynamic programming.
Read the full abstract
Dynamical Systems
Infinitesimal random dynamics of certain Veech groups on SU(2)-character varieties
Almost 20 years ago, the first and fourth authors found examples of SL(2,R)-invariant subbundles of Hodge bundles over Teichmüller curves having maximally degenerate Lyapunov spectrum. For these same surfaces, we show that a natural non-Abelian analogue has no zero Lyapunov exponents.
Read the full abstract
Functional Analysis
Boundedness of the averaging projections in nonlocally convex Lorentz sequence spaces and applications to basis theory
We study the boundedness of averaging projections associated with symmetric Schauder bases in quasi-Banach spaces. Although this property is standard in the Banach setting, it is far from clear in the absence of local convexity and, indeed, fails for a broad class of quasi-Banach spaces with a symmetric basis, including $\ell_p$ for $0<p<1$. Our main result shows that, nevertheless, the canonical basis of an entire class of weighted Lorentz sequence spaces, including the spaces $\ell_{p,q}$ for $0<q<1<p<\infty$, has uniformly bounded averaging projections. Thus, bounded averaging projections do not characterize local convexity among quasi-Banach spaces with symmetric bases. As applications, we obtain new consequences for the structure of special bases. In particular, as a byproduct of our approach, we derive new examples of conditional and almost greedy bases in nonlocally convex spaces.
Read the full abstract
Optimization and Control · Cross-listed from cs.LG
Quotient-Categorical Representations for Bellman-Compatible Average-Reward Distributional Reinforcement Learning
Average-reward reinforcement learning requires estimating the gain and a state-dependent bias, which solves the Poisson equation only up to an additive constant. Distributional analogues of the bias inherit this same ambiguity. We introduce a quotient-space formulation in which state-indexed laws are identified up to a common translation, together with a categorical parameterization that respects this symmetry. On this space, we define a projected average-reward distributional operator and show that it is well-defined, non-expansive in a coordinate Cramér metric, and admits fixed points. We interpret the projected fixed points through geometrically-stopped differential returns, whose categorical operators are contractive and whose fixed points converge in the Abel sense to a canonical fixed point of the projected undiscounted operator. We show that the means of the projected fixed points match the scalar bias up to an error caused by endpoint clipping of the categorical projections, and we bound this error. Finally, we establish almost-sure convergence and finite-iteration residual bounds for temporal-difference recursions with gain-centered rewards under i.i.d. and Markovian sampling. When the gain is unknown, we couple the recursion with an online gain estimator and prove the same guarantees under Markovian sampling, with a bound on the gain error.