Peer Revenue Returns subject to peer review

Papers

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

Mathematics

math Watch Mathematics (sign in to use)

Friday 20 March

Logic

Comment to "Almost disjoint sets, the dense set problem and the partition calculus"

This text highlights issues present in the proof of Lemma 6.10 of the Baumgartner (1943 – 2011) article "Almost disjoint sets, the dense set problem and the partition calculus" of 1976, and intends to present a correction at the same time it proves a stronger result mentioned in the article to have similar proof.
Read the full abstract

Analysis of PDEs

Stability and blow-up for a suspension bridge plate model with fractional damping and memory

We investigate a suspension bridge model described by a nonlinear plate equation incorporating internal fractional damping and infinite memory effects. The system also includes a nonlinear source term that may induce instability. Using semigroup theory, we first establish the local well-posedness of solutions in an appropriate energy space. We then derive conditions ensuring global existence and exponential stability of solutions. In contrast, when the initial energy is negative, we prove that solutions blow up in finite time, revealing a threshold phenomenon governing the long-term dynamics of the system. To complement the analytical results, we construct a numerical approximation using Summation-By-Parts finite differences with Simultaneous Approximation Terms (SBP-SAT) for spatial discretization and a Newmark scheme for time integration. The scheme preserves the structural properties of the continuous energy framework. Numerical experiments illustrate the stability and blow-up regimes predicted by the theoretical analysis.
Read the full abstract

All authors Iqra Kanwal, Jianghao Hao, Muhammad Fahim Aslam, Zayd Hajjej, Mauricio Sepúlveda-Cortés, Rodrigo Vejár-Asem

Functional Analysis

Hörmander's multiplier theorem on $H^p$-spaces in the rational Dunkl setting

On $\mathbb{R}^N$ equipped with a normalized root system $\mathcal R$ and a multiplicity function $k\geq 0$, let $dw(\mathbf x)=\Pi_{\alpha\in \mathcal R}|\langle \mathbf x,\alpha\rangle|^{k(\alpha)}\, d\mathbf x$, $\mathbf{N}=N+\sum_{\alpha\in \mathcal R}k(\alpha)$ denote the associated measure and the homogeneous dimension of the system $(\mathcal R,k)$ respectively. Let $\mathcal F$ stand for the Dunkl transform. For $0<p\leq 1$, let $m$ be a bounded function on $\mathbb{R}^N$, which satisfies the classical Hörmander's condition with smoothness $s>\mathbf{N}/p$. We show that the multiplier operator $\mathcal T_mf=\mathcal F^{-1}(m\mathcal Ff)$, initially defined on $H^p_{\mathrm{Dunkl}}\cap L^2(dw)$, has a unique extension to a bounded operator in $H^p_{\mathrm{Dunkl}}$, where the space $H^p_{\mathrm{Dunkl}}$ is defined by means of a Littlewood-Paley square function. To prove the theorem, we use special atomic and molecule characterizations of $H^p_{\mathrm{Dunkl}}$.
Read the full abstract

Dynamical Systems

Strong spectral gap for geometrically finite hyperbolic manifolds

Let $\Gamma < G := \operatorname{SO}(d+1, 1)$ for $d \geq 1$ be a Zariski dense, geometrically finite, discrete subgroup with critical exponent strictly greater than $d/2$. We show that $L^2(\Gamma\backslash G)$ admits a strong spectral gap, confirming a conjecture of Mohammadi and Oh. This extends the spherical spectral gap on $L^2(\Gamma\backslash \mathbb{H}^{d+1}) \cong L^2(\Gamma\backslash G/\operatorname{SO}(d+1))$, which follows by the works of Lax-Phillips, Patterson, and Sullivan by different methods. As a consequence, we establish rates of decay of matrix coefficients, and of exponential mixing of the frame flow, that are explicitly determined by the size of the strong spectral gap.
Read the full abstract

Dynamical Systems

Bifurcations for Lagrangian systems and geodesics I

This paper is Part I of a two-part series. We investigate bifurcation phenomena in Lagrangian systems with various boundary conditions and constraints, focusing on the interplay between Morse theory and the existence of multiple solutions through three principal configurations: Lagrangian trajectories connecting two submanifolds or with endpoints related by an isometry, and brake orbits in Lagrangian systems. For each configuration, we establish necessary and sufficient conditions for bifurcation using Morse index and nullity techniques, including classification of Rabinowitz-type alternative bifurcation scenarios. For Euler-Lagrange curves emanating perpendicularly from a submanifold, we develop a unified Morse-theoretic framework that rigorously connects geometric focal structure (e.g., conjugate points) and analytic bifurcation behavior (e.g., solution branching patterns).
Read the full abstract

Classical Analysis and ODEs

Evaluation of the symmetrized Mordell-Tornheim zeta function

In this paper we evaluate the symmetrized Mordell-Tornheim zeta function defined as \begin{equation*} \overline{\zeta}_n(w_1, \ldots, w_n) = \sum_{\substack{a_1, \ldots, a_n \in \mathbb{Z}^* \\ a_1 + \ldots + a_n = 0}} \frac{1}{\left| a_1^{w_1} \cdots a_n^{w_n} \right|} \end{equation*} where $n \ge 1$ is a positive integer representing the depth and $w_1, \ldots, w_n \ge 1$ are positive integers representing the weight $w = w_1 + \ldots + w_n$ of the function. Compared to the classical Mordell-Tornheim zeta function $\zeta_{MT,n}(w_1, \ldots, w_n; w_{n+1})$ which is restricted to the positive orthant (hyperoctant), the symmetrized one spans the entire $(n-1)$-dimensional hyperplane. We show that when the depth and the weight of the function are equal, that is for $\overline{\zeta}_n(1, \ldots, 1)$, it has a remarkably simple representation in terms of standard functions: \begin{equation*} \overline{\zeta}_n(1, \ldots, 1) = B_n(f^{(1)}(0), \ldots, f^{(n)}(0)) \end{equation*} where $B_n$ is $n$-th complete exponential Bell polynomial and $f^{(n)}(0)$ is $n$-th derivative at $x=0$ of function $f(x)$ defined as: \begin{equation*} f(x) = \ln \binom{-2x}{-x} \end{equation*} Additionally, we show the value can be expressed using the following polynomials with positive integer coefficients over the values of zeta function: \begin{equation*} \overline{\zeta}_n(1, \ldots, 1) = B_n(0, (2^2 - 2) \Gamma(2) \zeta(2), \ldots, (2^n - 2) \Gamma(n) \zeta(n)) \end{equation*} or equivalently, over the values of eta function: \begin{equation*} \overline{\zeta}_n(1, \ldots, 1) = B_n(0, 2^2 \Gamma(2) \eta(2), \ldots, 2^n \Gamma(n) \eta(n)) \end{equation*} The list of explicit values for small $1 \le n \le 10$ is available in the appendix.
Read the full abstract

Dynamical Systems

Similarity and superrigidity for transverse measured groupoids

Let $G$ be a locally compact, second countable, unimodular group and let $(X,\mu,Y)$ be an ergodic integrable transverse $G$-system. We show that $G \ltimes X$ endowed with the product measure and the restriction $(G \ltimes X)|_Y$ endowed with the transverse measure are similar ergodic groupoids. As a consequence, we prove a superrigidity phenomenon for higher rank transverse groupoids. More precisely, for $i=1,\ldots,k$, let $\mathbf{G}_i$ be a connected, simply connected, semisimple algebraic group over some local field $\kappa_i$ of characteristic zero. Let $G_i=\mathbf{G}_i(\kappa_i)$ be the $\kappa_i$-points of $\mathbf{G}_i$ and denote by $G=\prod_{i=1}^k G_i$. If we assume that $G$ has higher rank and each factor has positive rank, given an ergodic integrable transverse $G$-system $(X,\mu,Y)$, we prove that any unbounded Zariski dense morphism of the transverse groupoid $(G \ltimes X)|_Y$ into either an almost simple or a reductive algebraic group is superrigid.
Read the full abstract

Information Theory · Cross-listed from stat.AP

On the Limits of Prediction: Forecastability Profiles and Information Decay in Time Series

Forecasting accuracy is bounded by the information available about the future. This paper makes that statement precise using information-theoretic tools. Under logarithmic loss, the expected performance of any probabilistic forecast decomposes into two parts: an irreducible component and an approximation component. The irreducible term is the conditional entropy of the future given the available information, while the approximation term is the divergence between the true conditional distribution and the forecasting method. The gap between this conditional-entropy limit and an unconditional baseline is exactly the mutual information between the future observation and the declared information set. This leads to a definition of forecastability as the maximum achievable reduction in expected log loss. Evaluated across horizons, forecastability forms a profile that describes how predictive information varies with lead time. This profile reflects the dependence structure of the process and need not be monotone: predictive information may be concentrated at particular lags, including seasonal horizons, even when intermediate horizons contain little useful signal. From this profile, the paper defines the informative horizon set: the horizons at which forecastability exceeds a practical threshold. At horizons not in this set, the achievable gain over the unconditional baseline is necessarily small, regardless of the forecasting method used. The framework therefore separates what is learnable from what is not, and distinguishes limits imposed by the data from errors introduced by modelling. The result is a pre-modelling diagnostic that identifies where meaningful prediction is feasible before any model is chosen, providing a principled basis for allocating modelling effort across forecast horizons.
Read the full abstract

Quantum Algebra

Modular invariants and NIM-reps

Given a pivotal module category over a spherical fusion category, we introduce the encircling module, a module over the fusion algebra defined using the pivotal structure, and prove that it is isomorphic to the NIM-rep as a fusion algebra module. When applied to the $\mathcal{TM}$ realisation of the modular invariant partition function (arXiv:1911.09024), this identifies the NIM-rep multiplicities with the entries $Z_{I,I^\vee}$ of the modular invariant (its diagonal relative to the charge-conjugation invariant, which is the one given by the regular module category), providing a categorical generalisation of Böckenhauer, Evans and Kawahigashi's results (arXiv:math/9907149). We also show that for indecomposable module categories the dimension condition on $\mathcal{TM}$ required for modular invariance is automatically satisfied, and that $\mathcal{TM}$ is, as an algebra, the full centre of the module category in the sense of Davydov, Kong and Runkel (arXiv:1307.5956), which for module categories of algebras is the full centre of Fjelstad, Fuchs, Runkel and Schweigert (arXiv:hep-th/0612306). Consequently $\mathcal{TM}$ gives rise to Cardy algebras in the sense of Kong and Runkel (arXiv:0807.3356).
Read the full abstract

Differential Geometry

Computation of the multiplicities of zigzags

In this note, we explore various cohomological invariants on double complexes with the aim of finding their decomposition into irreducible parts, which are of square and zigzag shape. By studying the growth rate of the number of invariants given by the multiplicities of zigzags in the double complex of an n-dimensional complex manifold, we show that the De Rham, Dolbeault, Bott-Chern, Aeppli, and Varouchas cohomologies do not suffice to distinguish non-isomorphic double complexes. We also describe the zigzags counted by the Bigolin cohomology, and show how their dimensions are related to the multiplicities of odd zigzags. A special class of complex manifolds is given by the nilmanifolds. For a nilmanifold, the double complex of left-invariant forms is quasi-isomorphic to the double complex of differential forms. In dimension 6, we compute the double complex of forms of two nilmanifolds having the same Betti, Hodge and Bott-Chern numbers, but whose double complexes are non-isomorphic. We also compute the double complexes of a subclass of almost abelian nilmanifolds, which exist in any dimension.
Read the full abstract

Combinatorics

Lehmer Parking Functions and Their Outcomes

We introduce Lehmer parking functions and study their set of parking outcomes. Our main results establish that the number of outcomes of Lehmer parking functions of length $n$ is given by a Bell number, which is exactly the number of set partitions of an $n$ element set. We also show that the number of outcomes of weakly decreasing Lehmer parking functions is given by a Catalan number, which corresponds to a subset of set partitions on a set with $n$ elements referred to as non-intersecting set partitions.
Read the full abstract

All authors Melissa Beerbower, Jennifer Elder, Pamela E. Harris, Ilana Lavene, Lucy Martinez, Adam Martinson, Molly Oldham

Algebraic Geometry

Low degree points on singular plane curves

The purpose of this paper is to study low degree points on plane curves. We prove results analogous to those of Debarre and Klassen for singular plane curves with a finite number $\delta$ of ordinary nodes/cusps, where $\delta$ is bounded from above by a quadratic function in the degree of the plane curve.
Read the full abstract

All authors Zachary R. Canale, Nathan Chen, Zoe Curewitz, Jacob A. Daum, Karina Dovgodko, Carlos F. Santiago-Calderón, Shiv R. Yajnik

Information Theory · Cross-listed from cs.SE

Software Entropy: A Statistical Mechanics Framework for Software Testing

The notion of software entropy is often invoked to describe the tendency of software systems to become increasingly disordered as they evolve, yet existing approaches to quantify it are largely heuristic. In this work we introduce a formal definition of software entropy grounded in statistical mechanics, interpreting test suites as executable specifications, that is, as macroscopic constraints on the space of possible program implementations. Within this framework, mutation analysis provides a practical approximation of the locally accessible microstate space, allowing entropy-related quantities to be estimated empirically. We propose metrics that quantify how test suites restrict program space, including an information-weighted measure of the distribution of constraint power across tests. Applying these ideas to a real-world project, we show how test suites reduce software entropy and how information weights reveal structural differences in the contribution of individual tests that traditional metrics such as code coverage fail to capture.
Read the full abstract

Dynamical Systems

An Index Theorem in Relative K-Theory for First-Order Systems

Motivated by bifurcation of branches of homoclinic orbits of dynamical systems, we consider families of first-order equations on the real line and introduce a generalisation of previous index theorems by Pejsachowicz, and by Hu and Portaluri. The main novelties of our approach firstly concern the analytical setting, where we lift the common assumption that the equations are asymptotically hyperbolic. Secondly, we consider general compact parameter spaces instead of a single parameter, which results in a remarkably simple index formula in relative $K$-theory.
Read the full abstract

Optimization and Control · Cross-listed from cs.LG

Delightful Distributed Policy Gradient

Distributed reinforcement learning trains on data from stale, buggy, or mismatched actors, producing actions with high surprisal (negative log-probability) under the learner's policy. The core difficulty is not surprising data per se, but negative learning from surprising data. High-surprisal failures can dominate finite-batch updates through large perpendicular components, while high-surprisal successes reveal opportunities the current policy would otherwise miss. The Delightful Policy Gradient (DG) separates these cases by gating each update with delight, the product of advantage and surprisal, suppressing rare failures and preserving rare successes without behavior probabilities. In a tabular analysis, DG suppresses the perpendicular second moment of high-surprisal failures by a policy-overlap factor that vanishes as the learner improves. The advantage sign is essential for surprisal-based filtering: any learner-probability-only gate that suppresses rare failures also suppresses rare successes. On MNIST with simulated staleness, DG without off-policy correction outperforms importance-weighted PG with exact behavior probabilities. On a transformer sequence task with staleness, actor bugs, reward corruption, and rare discovery, DG often achieves nearly order-of-magnitude lower error. When all four frictions act simultaneously, its sample-efficiency advantage is order-of-magnitude and grows with task complexity.
Read the full abstract

Probability

A positive formula for volumes of moduli spaces of flat unitary connections on compact surfaces

We provide a manifestly positive expression for the volume of the moduli spaces of flat $\mathrm{U}(n)$-valued connections on punctured compact oriented surfaces. This volume is obtained by summing volumes of explicit polytopes describing coloured honeycombs on a polygon, in the spirit of the work of Knutson and Tao describing the spectrum of the sum of two hermitian matrices. As a corollary, we also provide a positive formula for marginals of the $\mathrm{U}(n)$-valued Yang-Mills measure on a compact oriented surface in terms of the probability distribution of an explicit path process.
Read the full abstract

All authors Quentin François, David García-Zelada, Thierry Lévy, Pierre Tarrago

Optimization and Control

Perturbation Duality for Robust and Distributionally Robust Optimization: Short and General Proofs

Duality is a foundational tool in robust and distributionally robust optimization (RO/DRO), underpinning both analytical insights and tractable reformulations. Whereas RO/DRO duality is commonly established through minimax arguments or conic duality, we use perturbation duality to obtain new, more general results with short proofs. We show that this perspective provides a natural and unifying framework for deriving RO/DRO dual formulations, proving the associated duality results, and diagnosing the regularity assumptions on which they depend. First, guided by perturbation duality, we establish new duality theorems for a recent DRO framework that unifies several canonical models, including $\phi$-divergence and Wasserstein models, through optimal transport subject to conditional moment constraints. Our results resolve an open conjecture on this DRO duality by clarifying the role of compactness: compactness itself is not necessary, but can be replaced by perturbation-based regularity conditions. Second, we revisit robust duality, commonly described as primal-worst equals dual-best. Using bifunctions, we unify dual-best formulations appearing in the literature and derive concise perturbation-based proofs that streamline recent results. Overall, the paper positions perturbation duality as a versatile and underutilized tool for RO and DRO, offering both conceptual unification and technical generality across a broad class of models.
Read the full abstract

Algebraic Geometry

Counting points on braid varieties and the Deligne–Simpson problem

We solve the isoclinic Deligne–Simpson problem for exceptional groups, completing a program initiated by Sage et al. and Jakob–Yun. As a by-product, we obtain new examples of physically rigid irregular connections on the projective line. Our approach uses the Riemann–Hilbert correspondence to reduce the problem to determining the non-emptiness of certain braid varieties associated to periodic braids. We show that this can be achieved by counting points over finite fields. Our approach is inspired by Lusztig's construction of a map from conjugacy classes in the Weyl group to unipotent classes.
Read the full abstract

Differential Geometry

Optimal Transport and Generalized Lagrangian Mean Curvature Flows on Kim-McCann Metrics

We express the mean curvature flow of Lagrangian submanifolds in pseudo-Riemannian manifolds endowed with the Kim-McCann-Warren metric within the framework of generalized mean curvature flow on Kim-McCann manifolds. While generalized mean curvature flow has been studied in Kähler geometry, our work shows that techniques from para-Kähler geometry arise naturally in the Kim-McCann setting. Using this perspective, we prove that the Lagrangian condition is preserved along the flow. By identifying generalized mean curvature flow with Lagrangian mean curvature flow, we show that the Ma-Trudinger-Wang regularity theory applies to this setting. In particular, the cross-curvature positivity condition of Kim-McCann yields smoothly converging flows of Lagrangian submanifolds. Under the cross-curvature condition, any Lagrangian submanifold avoiding the cut locus converges exponentially to a stationary submanifold, which locally arises as the graph of an optimal transport map. Our framework substantiates the analogy between special Lagrangian geometry in almost Calabi-Yau manifolds and optimal transport theory in the Kim-McCann setting. In particular, we show that Kim-McCann manifolds equipped with a para-holomorphic volume form serve as the natural counterpart to almost Calabi-Yau manifolds.
Read the full abstract

Combinatorics

Partition regularity in imaginary quadratic rings of integers

We obtain partition regularity results for homogeneous quadratic equations whose parametrized solutions admit nice factorizations into linear forms over rings of integers of imaginary quadratic fields. To do so, we develop number-theoretic results of independent interest on such fields, such as a characterization for aperiodic completely multiplicative functions, the Turán-Kubilius inequality, and a new concentration estimate for multiplicative functions.
Read the full abstract

All authors Sebastián Donoso, Andreu Ferré Moragues, Andreas Koutsogiannis, Wenbo Sun

Rings and Algebras

Central Products of Cayley-Dickson Loops

This paper studies the triviality of commutators in central products of Cayley-Dickson loops. Two immediate outcomes of this study are (1) the construction of a sequence of non-commutative di-associative loops in which the probability that a random commutator is trivial approaches 1, and (2) an easy proof that if two central products of $n$-fold Cayley-Dickson loops are isomorphic for $n\geq 3$, then the loops in the first product are term-wise isomorphic to the loops in the second product.
Read the full abstract

Geometric Topology

Every Weak Perron Number is an End-Periodic Stretch Factor

Given any weak Perron number $\lambda$, we construct an end-periodic homeomorphism $f:\Sigma\rightarrow \Sigma$ with Handel-Miller stretch factor equal to $\lambda$ where $\Sigma$ is a connected infinite-type surface with finitely many ends all accumulated by genus.

Mathematical Physics · Cross-listed from gr-qc

Singular structures and causality of the Schwarzschild Green's function in the frequency domain

We study two singular spectral components of the Green's function of a Schwarzschild black hole and their interpretation in the frequency domain: (i) the low-frequency branch cut, which yields corrections to Price's law tails in the form of inverse power laws weighted by logarithmic terms; and (ii) the quasinormal-mode spectrum, which generates a redshifted response for sources extended toward the horizon. We show that the frequency-domain Green's function can be naturally interpreted in terms of greybody factors, providing the first analytical justification for recent phenomenological ringdown models based on these quantities. For sources localized outside the peak of the potential barrier, we identify two tail contributions activated with a time delay, arising from backscattering of the prompt response and of the ringdown signal. We show that corrections to Price's law can be relevant at intermediate times, when the ringdown still dominates the waveform. For sources localized inside the potential barrier peak, the tail is suppressed and the signal is instead dominated by quasinormal frequencies. In this regime, these spectral components produce both the ordinary quasinormal-mode ringdown and an infinite tower of exponentially decaying terms governed by the horizon surface gravity, the so-called redshift terms. We demonstrate that this component is not screened by geometric features of the background spacetime and persists up to late times, as supported by numerical investigations of perturbative waveforms. Our results provide a mathematical foundation for phenomenological modeling of the branch-cut contribution at intermediate times, which is relevant for prospective observations of tails, and strong evidence for the presence of redshifted components from intermediate to late times.
Read the full abstract

Numerical Analysis

Convergence Analysis of the Random Bisection Method

We propose a generalized version of the bisection method where the cutting point between the two subintervals is chosen at random following an arbitrary distribution. We compute expected convergence rates with respect to any arbitrary a priori distribution for the position of the root in the initial interval and proved that it depends only on the the expectation $\mathbb{E}[c(1-c)]$ of the cut $c$. We also provide a generalization of the method for $K$ random cuts and study its convergence properties. Most probabilistic derivations are kept fairly simple for the ease of understanding of a larger audience. Our theoretical results are then validated numerically using statistical simulation.
Read the full abstract

General Topology

On exact capacities

We consider capacity (fuzzy measure, non-additive probability) on a compactum as a monotone cooperative normed game. We introduce topological analogues of well known class of exact games and show that these classes form subfunctors of the capacity functor which lie between known subfunctors of convex capacities and balanced capacities. It is natural to consider probability measures as elements of core of such games. We describe exact capacities as envelopes of the convex closed sets of probability measures. Using such representation we prove openness of the functor of exact capacities. We also consider strongly exact capacities and pose the problem of coincidence of these two classes.
Read the full abstract

Analysis of PDEs

Half Strong Ill-Posedness of $2 \frac{1}{2}$D Electron Magnetohydrodynamics with Fractional Resistivity

We study the $2\frac{1}{2}$D electron magnetohydrodynamics (MHD): the electron MHD system that has $3$D magnetic field but is independent of $z$-variable. We establish a "half" strong ill-posedness result in $2\frac{1}{2}$D electron MHD with fractional resistivity $(-\Delta)^\alpha$ in the supercritical Sobolev space $H^{\beta}\times H^{\beta-1}$ for $3<\beta<4-2\alpha$. Specifically, we construct small initial data $(a_0,b_0)$ whose solution develops a norm inflation in $a$ but the norm of $b$ remains small.
Read the full abstract

Dynamical Systems · Cross-listed from stat.ME

Goal-oriented learning of stochastic differential equations using error bounds on path-space observables

Stochastic differential equations (SDEs), which serve as the governing equations for dynamical systems in a broad range of applications, can become cost-prohibitive for numerical simulation at scales necessary for quantifying key properties. Surrogate models of the drift function of an SDE, learned from data of the high-fidelity system, are routinely used to increase the efficiency of simulation and prediction of properties. However, standard choices of loss function for learning the surrogate model fail to provide error guarantees in certain path-dependent observables, such as transition times. This paper introduces an error bound for path-space observables and employs it as a novel variational loss for the goal-oriented learning of the drift function of a SDE. We show the error bound holds for a broad class of observables, including mean first hitting times on unbounded time domains. We derive an analytical gradient of the goal-oriented loss by leveraging the formula for Fréchet derivatives of expected path functionals, which remains tractable for implementation in stochastic gradient descent schemes. We demonstrate that surrogate models of overdamped Langevin systems developed via goal-oriented learning achieve improved accuracy in predicting the statistics of a first hitting time observable and robustness to distributional shift in the data.
Read the full abstract

Optimization and Control

Time-Reversed BSDEs for Accurate Gradient Estimation in Diffusion Models

There is a growing literature adopting a stochastic optimal control (SOC) perspective to fine-tune diffusion models and related generative policies. A prominent class of methods, known as iterative diffusion optimization, solves the SOC problem by simulating the diffusion process, evaluating a loss function, and applying stochastic optimization algorithms, with adjoint matching emerging as a state-of-the-art approach. However, the adjoint process used in these methods is not adapted to the forward diffusion filtration, which can lead to unstable or high-variance gradient estimates. In this paper, we revisit gradient estimation in diffusion models through the lens of backward stochastic differential equations (BSDEs). We propose an alternative estimator based on a time-reversed BSDE formulation introduced in our prior work, which produces an adjoint process adapted to the underlying filtration. This adapted structure leads to more stable gradient estimates with potentially lower variance. We analyze the accuracy of the proposed estimator and compare it with adjoint matching. Numerical experiments on fine-tuning toy diffusion models demonstrate improved gradient stability and competitive performance.
Read the full abstract

Analysis of PDEs

Data assimilation for the barotropic Navier-Stokes system

We consider a continuous data assimilation method for the barotropic Navier–Stokes system. The observed solution is supposed to be bounded on the whole time period of observation, while the synchronized solution, usually provided by a numerical method, belongs to the class of dissipative solutions that is considerably larger than the class of conventional weak solutions. A complete synchronization is shown on any compact prediction interval provided the nudging parameters are chosen appropriately.
Read the full abstract

Optimization and Control

On the Gap Between H2 Optimal Control and Disturbance Decoupling

We study the relationship between disturbance decoupling (DD) and H2 optimal control for linear time-invariant (LTI) systems, revealing a fundamental gap between DD subspace constraints and semi-definite program (SDP)-based H2 minimization. We show that DD is equivalent to the existence of zero H2 gain without requiring internal stability, whereas SDP-based H2 minimization strictly optimizes over stabilizing controllers and therefore fails to recover DD controllers when the closed-loop dynamics may be marginally stable. Moreover, we show that the trace representation of H2 norms further biases solutions away from complete DD. Motivated by this, we formulate a bilinear matrix inequality (BMI)-constrained optimization program that directly enforces the DD subspace condition to compute DD controllers. We propose a difference-of-convex (DC) iterative algorithm that preserves DD and stability at every iteration, and establish its convergence to Karush-Kuhn-Tucker (KKT) points under standard constraint qualification conditions. Numerical experiments on a four bus power network demonstrate that the proposed algorithm achieves significantly better disturbance rejection while enabling optimization of additional performance metrics. The resulting framework establishes a computationally tractable link between geometric DD theory and optimization-based controller design.
Read the full abstract