Optimization and Control · Cross-listed from cs.CG
Towards Scalable Persistence-Based Topological Optimization
Abderrahim Bendahi, Alexandre Duplessis and Arnaud Fickinger
Persistence-based topological optimization deforms a point cloud $X \subset \mathbb{R}^d$ by minimizing objectives of the form $L(X) = \ell(\mathrm{Dgm}(X))$, where $\mathrm{Dgm}(X)$ is a persistence diagram. In practice, optimization is limited by two coupled issues: persistent homology is typically computed on subsamples, and the resulting topological gradients are highly sparse, with only a few anchor points receiving nonzero updates. Motivated by diffeomorphic interpolation, which extends sparse gradients to smooth ambient vector fields via Reproducing Kernel
Hilbert Space (RKHS) interpolation, we propose a more scalable pipeline that improves both subsampling and gradient extension. We introduce subsampling via random slicing, a lightweight scheme that promotes iteration-wise
geometric coverage and mitigates density bias. We further replace the costly kernel solve with a fast Nadaraya-Watson (NW) Gaussian convolution, producing a globally defined smooth update field at a fraction of the computational cost, while being more suited for topological optimization tasks. We provide theoretical guarantees for NW smoothing, including anchor approximation bounds and global Lipschitz estimates. Experiments in $2$D and $3$D show that combining random slicing with NW smoothing yields consistent speedups and improved objective values over other baselines on common persistence losses.
Read the full abstract ▾
Dynamical Systems
An elegant model of the geodesic flow on the modular surface
Pierre Arnoux and Thomas A. Schmidt
Caroline Series' [The modular surface and continued fractions, J. Lond. Math. Soc. (2), 31, no.~1, (1985), 69–80] gives a clear framework linking, in a deceptively simple way, the dynamics of the geodesic flow on the modular surface with the dynamics of the regular continued fraction, through a well-chosen symbolic coding. It has been called required reading for those interested in the symbolic dynamics of geodesic flows, and has had consequences in symbolic dynamics, ergodic theory,
hyperbolic geometry, and continued fraction theory. In this overview, we give an indication of why this is so, sketch some of the history related to the paper, and also point to some later works.
Read the full abstract ▾
Combinatorics
Average Hitting Times and Recurrence Structures I: Powers of Cycle Graphs
Tsuyoshi Miezaki and Shunya Tamura
We investigate the average hitting times of simple random walks on the $k$-th power graph $C_N^k$ of the cycle graph $C_N$. First, we show that the average hitting times are characterized by a difference equation corresponding to the graph
Laplacian. Next, by using the cyclic
symmetry of $C_N^k$, we derive a spectral representation via
Fourier analysis. Furthermore, by applying factorization and partial fraction decomposition of the corresponding difference operator, we obtain an explicit formula for the average hitting times consisting of a quadratic term and finitely many correction terms. These correction terms are described by second-order linear recurrence sequences associated with the characteristic
polynomials, and can be regarded as natural generalizations of Fibonacci-type sequences. As a consequence, our formulas recover the known results for cycle graphs and squares of cycle graphs in a unified way. Moreover, from the formulas obtained for average hitting times, we derive explicit formulas for the effective resistances, the numbers of spanning trees, the numbers of two-component spanning forests, and the numbers of spanning trees of vertex-identified graphs. In particular, for the third power graph $C_N^3$ of the cycle graph, all of these quantities are written explicitly in terms of complex conjugate Fibonacci-type sequences. Our results clarify structural relations between random walk quantities and combinatorial quantities on cycle power graphs.
Read the full abstract ▾
Optimization and Control
Localization for nonlocal gradient-based optimal control problems
Javier Cueto and Joshua M. Siktar
In this paper we consider optimal control problems in the nonlocal function space framework of Bellido-2023, where there are two different parameters: a horizon parameter $\delta > 0$; and a fractional parameter $s \in (0, 1)$. The constraints are given in the form of minimizing an energy density, and we will focus on two particular cases: the well-posed case where the underlying energy density is convex and is given by the nonlocal $p$-
Laplacian; and a more general poly/quasiconvex energy for which minimizers exist but may not be unique. The study is concluded by analyzing the approximation to local problems in two parallel ways, either taking the fractional parameter $s$ to $1$ or the horizon parameter $\delta$ to $0$.
Read the full abstract ▾
Combinatorics
No-three-in-line sets on the checkerboard grid
Thomas Prellberg
The classical no-three-in-line problem asks for the largest number $D(n)$ of points that can be chosen from an $n\times n$ grid with no three collinear; it remains open whether the elementary upper bound $2n$ is always attainable. We study a checkerboard-restricted variant in which all chosen points are monochromatic under the colouring of the grid by the parity of $x+y$. If $D_{\mathrm{mono}}(n)$ denotes the largest number of monochromatic points with no three collinear, the monochromatic diagonals already give $D_{\mathrm{mono}}(n)\le 2n-2$. The main object of the paper is a four-direction linear-programming relaxation on a fixed colour class, using rows, columns and the two diagonal families of slopes $\pm1$. For the ordinary square-grid problem this relaxation gives the trivial bound; on the checkerboard it is substantially tighter, and for $2\le n\le 16$ its floor agrees with the exact single-colour optimum except at four side lengths, where the gap is one. After
symmetry reduction the dual relaxation has three one-dimensional reduced forms, according to the parity of $n$ and the chosen colour class. The central construction is an exact continuum dual certificate for the continuum problem associated with the scaled
symmetry-reduced odd-fat case: explicit nonnegative functions $A$ and $B$ satisfying the continuum obstacle inequalities, with objective value the middle real root $\alpha\approx1.5768$ of $401\alpha^3-1744\alpha^2+2240\alpha-768=0$. Combined with a continuum-to-discrete sampling
theorem of Aujla, Prellberg and Sandhu, this certificate yields the asymptotic upper bound $D_{\mathrm{mono}}(n)\le\alpha n+O(1)$. Finite LP computations are consistent with $\alpha$ as the exact limiting slope of the relaxation, and the exact small-$n$ data suggest, more speculatively, that the true checkerboard optimum tracks the same scale.
Read the full abstract ▾
Information Theory · Cross-listed from cs.LG
Fast Rates for Offline Contextual Bandits with Forward-KL Regularization under Single-Policy Concentrability
Qingyue Zhao, Kaixuan Ji, Heyang Zhao et al. (+1 more)
Kullback-Leibler (KL) regularization is ubiquitous in reinforcement learning
algorithms in the form of reverse or forward KL. Recent studies have demonstrated $\epsilon^{-1}$-type fast rates for decision making under reverse KL regularization, in contrast to the standard $\epsilon^{-2}$-type sample complexity. However, for forward-KL-regularized objectives, existing statistical analyses are either not applicable or result in $\tilde{O}(\epsilon^{-2})$ slow rates. We take the first step towards addressing this problem via a streamlined analysis of forward-KL-regularized offline CBs. We give the first $\tilde{O}(\epsilon^{-1})$ upper bounds in tabular and general function approximation settings, both under notions of single-policy concentrability. In particular, our convex-analytical pipeline unifies these settings by exploiting the pessimism principle in a novel way and completely bypasses the proof routines in previous works based on the mean value
theorem, which might be of independent interest. Moreover, we provide rate-optimal lower bounds, manifesting the tightness of our upper bounds in terms of statistical rates. Our lower bounds also demonstrate that the forward-KL-regularized sample complexity recovers the unregularized slow rate in the low-regularization regime, similarly to the reverse-KL regularization.
Read the full abstract ▾
All authors Qingyue Zhao, Kaixuan Ji, Heyang Zhao, Quanquan Gu
Analysis of PDEs
Kinetic theory for Transformers and the lost-in-the-middle phenomenon
Mitia Duerinckx, Borjan Geshkovski and Stefano Rossi
We study causal self-
attention dynamics – a toy model for decoder
Transformers – which we interpret as a non-exchangeable interacting particle system. Adapting cumulant expansions to the triangular causal dependency structure of the model, and appealing to non-hierarchical methods to estimate correlations using Glauber calculus, we prove a quantitative mean-field limit result and a next-order characterization of correlations. For iid uniformly distributed
tokens, the limiting correlation equation can be solved in closed form and we obtain a rigorous explanation of the empirically observed lost-in-the-middle phenomenon: the
token retrieval profile, as a function of the source position in the
prompt, is $\mathsf{U}$-shaped, with primacy, recency, and a unique interior minimum under an explicit smallness condition.
Read the full abstract ▾
Numerical Analysis
A Variational Equation and Lower Bound for the Linear Least-Squares Backward Error
Eric Hallman
This paper derives a new variational equation for the linear least-squares backward error by expressing the backward error in terms of a generalized
eigenvalue problem and using results from indefinite linear
algebra. For problems with multiple right-hand sides, the variational equation also shows that the backward error can be decomposed as a sum of smaller backward error problems. Applications to stopping criteria for iterative methods are considered, and a new sketching-based lower bound is proposed which is provably of quality comparable to the sketched Karlson-Waldén estimate.
Read the full abstract ▾
Algebraic Geometry
Generic vector fields on isolated complex hypersurface germs
Diogo da Silva Machado and José Seade
At a smooth point, a generic holomorphic vector field vanishing there has index one. At an isolated hypersurface singularity, the corresponding generic GSV index is an invariant of the analytic germ. We study this invariant, denoted by $K$, and prove the sharp bound $K\ge1+(-1)^n\tau$, where $n$ is the dimension and $\tau$ the Tjurina number. We characterize equality by weighted homogeneity in odd dimensions; Liu and Zhang subsequently established the converse in even dimensions. Beyond equality, $K$ detects analytic information not determined by the Milnor and Tjurina numbers. We compute it on the normalization of a reduced plane curve and interpret its difference from the Euler characteristic of the Milnor fibre as a logarithmic Poincaré–
Hopf index on a resolution. This gives a
geometric explanation for a striking contrast: the difference is nonnegative for curves, but can be arbitrarily negative for surfaces. Our double-suspension
theorem, $K(f+z^2+w^2)=K(f)$, carries these phenomena to every odd and every even dimension, respectively. In every dimension we construct singularities whose generic index exceeds the sharp bound by exactly two. We
conjecture that an excess of one is impossible.
Read the full abstract ▾
Optimization and Control
Select-then-differentiate: Solving Bilevel Optimization with Manifold Lower-level Solution Sets
Saeed Masiha, Zebang Shen, Negar Kiyavash et al. (+1 more)
We study optimistic bilevel optimization when the lower-level problem has a non-isolated
manifold of minimizers. In this setting, the hyper-objective may be non-differentiable because the upper-level criterion must choose among multiple lower-level solutions. Under a local Polyak–Łojasiewicz (PŁ) condition, we show that differentiability does not require the lower-level solution set to be a singleton: uniqueness of the optimistic selection is sufficient. This yields an explicit pseudoinverse-based hyper-gradient formula extending the classical singleton-minimizer result. We further characterize the regularity of the hyper-objective: non-degeneracy of the selected minimizer along the solution
manifold yields local smoothness, while failure of uniqueness can create many non-differentiable points and failure of non-degeneracy can destroy all positive Hölder regularity of the hyper-gradient. Motivated by this theory, we propose HG-MS, a select-then-differentiate method combining explicit optimistic selection with efficient pseudoinverse-based hyper-gradient computation. Despite the nonconvex nature of optimistic selection over the lower-level solution
manifold, we show that HG-MS converges to a stationary point of the optimistic objective with complexity governed by the intrinsic dimension of the solution
manifold rather than its ambient dimension. Empirically, we test a practical variant of HG-MS for matched-budget LLM source reweighting. This variant preserves the select-then-differentiate principle and obtains the best GSM8K/MATH scores across the tested backbones, along with competitive or best MT-Bench instruction-following results.
Read the full abstract ▾
All authors Saeed Masiha, Zebang Shen, Negar Kiyavash, Niao He
Optimization and Control
An optimal control problem for Stokes-Cahn-Hilliard-Oono equations with regular potential
Arghya Kundu
This article discusses an optimal control problem for a phase field model of two immiscible incompressible fluid flow, incorporating surface tension effects. The optimal control problem is defined with a $L^2$-cost functional and subject to the constraints governed by a system of coupled Stokes-Cahn-Hilliard-Oono equations. In this model, fluids are separated by a dynamic diffuse interface of finite width. We investigate the optimality condition of a given control. Initially, we establish the existence of an optimal solution for the coupled optimal control problem. Subsequently, we derive the optimality condition with respect to the corresponding adjoint system.
Read the full abstract ▾
Numerical Analysis
An NPDo Approach for Principal Joint SVD-type Block Diagonalization
Ren-Cang Li, Li Wang and Mei Yang
This paper is concerned with partial Joint SVD-type Block Diagonalization of several
matrices so that the extracted diagonal parts collectively optimally assume part of the total mass of all given
matrices. For that reason, it will be referred also as Principal Joint SVD-type Block Diagonalization. When each block-size is 1-by-1, it is about finding a dominant partial joint SVD decomposition for the
matrices of interests. An NPDo approach is proposed for maximizing the common dominant block-diagonal parts collectively. It is shown that the NPDo approach combined with Gauss-Seidel-type updating is globally convergent to a stationary point while the objective increases monotonically. Numerical experiments are presented to illustrate the efficiency of the NPDo approach.
Read the full abstract ▾
Probability
Fluctuations for the critical free Bose gas
Quirin Vogel
We study the critical free Bose gas from a probabilistic perspective with a focus on the three-dimensional case. We show that, unlike in the subcritical and supercritical regime, the critical asymptotics depend on the second heat-trace coefficient $a_1$ and hence on the geometry and boundary conditions. For $a_1<0$, long loops occur on the scale $L^2\log L$ and yield Poisson–Dirichlet statistics; for $a_1<0$, such loops are suppressed and the relevant fluctuations are produced by the short-loop bulk. We also derive asymptotics for the partition function, the one-particle reduced density
matrix, and loop statistics. The fluctuation limits are governed either by a regularized Fredholm determinant of the Green operator or by Gaussian local central limit
theorems, depending on the boundary regime.
Read the full abstract ▾
Differential Geometry
On centerless unimodular contact Lie algebras
Agustín Garrone
We provide an elementary proof that, in a (transversely) unimodular contact Lie
algebra, the adjoint action of the Reeb vector is nilpotent except when the Lie
algebra is isomorphic to either $\mathfrak{sl}(2,\mathbb{R})$ or $\mathfrak{su}(2)$. We introduce a class of contact Lie
algebras, called DS-contact Lie
algebras, containing all K-contact Lie
algebras, and deduce from the previous result that the only centerless unimodular examples in this class are precisely $\mathfrak{sl}(2,\mathbb{R})$ and $\mathfrak{su}(2)$. This gives an alternative proof of the previously known fact that centerless unimodular Sasakian Lie
algebras are isomorphic to either $\mathfrak{sl}(2,\mathbb{R})$ or $\mathfrak{su}(2)$. Some other results known to hold for Sasakian Lie
algebras are generalized as well. We investigate several properties of DS-contact Lie
algebras in relation to Frobenius Lie
algebras, and also classify them in dimension five. Some implications for the contact Lefschetz condition are explored.
Read the full abstract ▾
Functional Analysis
Characterizations of the UMD property via tail estimates for tangent processes
Gergely Bodó and Ivan Yaroslavtsev
We characterize the UMD property of a
Banach space by tail inequalities for maximal functions of tangent conditionally symmetric processes. More precisely, we prove that a
Banach space $V$ is UMD if and only if for some (equivalently, for all) $p\in(0,\infty)$ one has that \[ \mathbb P(\sup_{r\geq 0} \| N_r\|>t)\lesssim_{p,V}\Bigl(\frac{s^p}{t^p}+\mathbb P(\sup_{r\geq 0} \| M_r\|>s)\Bigr), \qquad s,t>0, \] for all tangent conditionally symmetric $V$-valued processes $M$ and $N$. We further show that this estimate is equivalent to suitable Lorentz norm inequalities for the associated maximal functions, and obtain analogous characterizations in the discrete-time, continuous-time, and purely discontinuous settings.
Read the full abstract ▾
Optimization and Control · Cross-listed from cs.LG
Navigating LLM Valley: From AdamW to Memory-Efficient and Matrix-Based Optimizers
Aditya Ranganath
Training large
language models requires optimization
algorithms that are not only statistically effective, but also computationally and
memory efficient at extreme scale. Although Adam remains the dominant optimizer for large-scale
language-model pretraining and fine-tuning, recent work has revisited nearly every component of the optimization stack: adaptive moment estimation, decoupled weight decay,
memory footprint,
curvature approximation, sign-based updates, large-batch stability, low-rank gradient structure, and
matrix-wise orthogonalized updates. This survey reviews optimizer design for large
language models through a systems-and-optimization lens. We organize the literature into classical first-order optimizers, adaptive optimizers,
memory-efficient variants, second-order and
curvature-aware methods, sign-based and discovered optimizers, low-rank and projection-based methods, and
matrix-based optimizers such as
Muon. We also discuss benchmarking methodology, including hyperparameter
fairness, scale dependence, wall-clock efficiency,
token efficiency,
memory overhead, and downstream evaluation. We argue that optimizer research for LLMs is entering a new phase: moving from single-
algorithm speedup claims toward rigorous, scale-aware comparisons that jointly evaluate convergence, stability,
memory, and implementation complexity.
Read the full abstract ▾
Analysis of PDEs
On singular problems in nonreflexive fractional Orlicz-Sobolev spaces
Marcos L.M. Carvalho, Luana C. M. Lima, Carlos A.P. Santos et al. (+1 more)
In this work, we deal with existence and uniqueness of positive solution $u_s$ for the singular quasilinear problem $(-\Delta_{\Phi})^su=u^{-\gamma}$ in the nonreflexive fractional Orlicz-Sobolev $ W^{s}_0L^{\Phi}(\Omega)$ for $0<s<1$. Furthermore, we show that $u_s$ converges in $L^{\Phi}(\Omega)$ to the unique positive solution $u\in W^{1}_0L^{\Phi}(\Omega)$ of the problem $-\Delta_{\Psi}u=u^{-\gamma}$ as $s \uparrow 1$, where $\Psi$ is an appropriate $N$-function equivalent to the $N$-function $\Phi$. The main difficulties to obtain existence of weak solutions for both singular quasilinear problems are that their associate energy functionals may not be well-defined on their whole natural workspaces due to the lack of the reflexivity and the presence of the singular term. To overcome these difficulties, we will use the minimization method and present a new approach to building appropriate test functions to prove that the problems have positive minimizers that we showed to be weak solutions of them, respectively.
Read the full abstract ▾
All authors Marcos L.M. Carvalho, Luana C. M. Lima, Carlos A.P. Santos, Maxwell L. Silva
Number Theory
Rational points on smooth surfaces in $\mathbb{P}^3$ over finite fields
Yves Aubry and José Felipe Voloch
We improve a bound due to the second author on number of rational points on smooth surfaces in $\mathbb{P}^3$ over finite fields and look at families of surfaces that achieve or nearly achieve this bound, for which we compute their exact number of rational points. These computations may have independent interest.
Read the full abstract ▾
Optimization and Control
A Certificate of Unboundedness for Polynomial Optimization Problems
Rohan Rele and Angelia Nedich
Global
polynomial optimization methods typically rely on compactness of the feasible region in order to find solutions. These methods can incur considerable computational expense and most commercially available solvers do not verify the existence of a solution prior to undergoing global search. In this manuscript we propose a simple pre-processing
algorithm to determine if an arbitrary
polynomial optimization problem is unbounded from below thereby providing information about the problem's asymptotic geometry prior to solving the problem if a solution can be found.
Read the full abstract ▾
Logic
On the Intermediate Models of Strongly Compact Prikry Forcing
Tom Benhamou, Sebastiano Thei and Ben-Zion Weltsch
We analyze the intermediate models of the strongly compact Prikry forcing. We exhibit a simple combinatorial property which, for a given supercompact cardinal $\kappa$, characterize the projections of all projections of the strongly compact Prikry forcing using $\kappa$-complete fine measures. Considering level-by-level results, if $\kappa$ is $2^\lambda$-strongly compact, we characterize the forcings of size $\leq\lambda$ which are projections of that $\lambda$-strongly compact Prikry forcing. Our characterization generalizes several known results, including those of Benhamou-Hayut-Gitik and folklore results regarding the class of $\kappa$-distributive forcing notions which are embedded into the supercompact Prikry forcing. Fixing a $\kappa$-complete fine measure $\mathcal{U}$ on $P_\kappa(\lambda)$, we also provide Rudin-Keisler like critiria for the existence projections from the strongly compact Prikry forcing with $\mathcal{U}$. Finally, we prove that among all projections of the $\lambda$-strongly compact Prikry forcing, the class of forcings of cardinality $\lambda$ are exactly those for which there is a projection map which depends only on the stem of the Prikry condition. We also give partial results regarding projections of arbitrary cardinality.
Read the full abstract ▾
Algebraic Geometry
A curve and its abstract generalized Jacobian
Benjamin Castle, Ishai Dan-Cohen and Assaf Hasson
To a smooth proper curve $C$ over a field $k$ equipped with a $k$-point $c$ and an effective divisor $\mathfrak m$ coprime to $c$, one may associate the abstract group $J_{\mathfrak m}(\bar k)$ of $\overline k$-points of the generalized Jacobian, as well as a subset \[ \tag{*} \big(C\setminus \operatorname{Supp}(\mathfrak m)\big)(\bar k) \subset J_{\mathfrak m}(\bar k). \] We show that the data $(C,c,\mathfrak m)$ can be retrieved from (*) up to a twist by an
automorphism of $\overline k$, proving a
conjecture of Booher and Voloch. By a result of Booher and Voloch this shows that when $k$ is a finite field, the same data may also be retrieved from $L$-functions of characters of certain Galois extensions of the function field of $C$. The proof is a generalization of Zilber's well known work "A curve and its abstract Jacobian".
Read the full abstract ▾
Commutative Algebra
Quadratic linear strands of prime ideals
Giulio Caviglia and Alessandro De Stefani
We prove sharp estimates on the quadratic strand of the resolution of any homogeneous
prime ideal in a standard graded
polynomial ring over an arbitrary field. Our bounds only depend on the height of the
prime ideal, and they are optimal since for every $h \geq 1$ we show that there exists a
prime ideal of height $h$ achieving them. In particular, we show that a
prime ideal of height $h$ can contain at most $h^2$ quadratic minimal generators, and that there exists a
prime ideal minimally generated by $h^2$ quadrics.
Read the full abstract ▾
Category Theory
A categorical description of simple Beth companions
Luca Carai, Miriam Kurtzhals and Tommaso Moraschini
A pp expansion of a quasivariety $\mathsf{K}$ is said to be simple when it is of the form $\mathsf{K}[\mathscr{L}_\mathcal{F}]$. For instance, when $\mathsf{K}$ has the amalgamation property, all its pp expansions are simple. It is shown that the simple pp expansions of a quasivariety $\mathsf{K}$ coincide with the quasivarieties $\mathsf{M}$ for which the forgetful functor $U \colon \mathsf{M} \to \mathsf{K}$ is well defined and induces an isomorphism from $\mathsf{M}$ to a mono-reflective subcategory of $\mathsf{K}$. As a consequence, if a quasivariety $\mathsf{K}$ possesses a simple Beth companion $\mathsf{M}$, then $\mathsf{M}$ is the unique (up to term equivalence) quasivariety whose monomorphisms are regular that, moreover, satisfy the categorical description of simple pp expansions of $\mathsf{K}$ given above.
Read the full abstract ▾
Information Theory · Cross-listed from quant-ph
Enhanced quantum capacity thresholds from symmetry
Avantika Agarwal, Amolak Ratan Kalra, Sungjai Lee et al. (+4 more)
The
quantum capacity captures the value of a
quantum channel for transmitting
quantum information, establishing the fundamental limits on
quantum communication. In spite of its central role in
quantum information theory, the
quantum capacity of most channels is unknown, with wide gaps between the best upper and lower bounds. Even deciding whether a channel has nonzero capacity – finding its capacity threshold – is difficult. In this paper we report significant increases in the capacity thresholds of two prototypical noise models: the depolarizing channel and Pauli channels. In the case of the depolarizing channel, this is the first improvement in 18 years, giving a bigger increase beyond the hashing bound than all previous improvements combined. Our starting point is the representation theoretic framework recently proposed by Bhalerao and Leditzky (2025) to compute coherent information for special permutation invariant states. We generalize their framework to the full symmetric subspace, which allow us to optimize coherent information over rank two states in that space. A representation theoretic calculation shows that exponentially many Kraus operators of the channel annihilate the symmetric space, corresponding to a massive decrease in environment
entropy for states on the symmetric space compared to the maximally mixed state. This explains the enhanced coherent information as a manifestation of degeneracy for the resulting codes.
Read the full abstract ▾
All authors Avantika Agarwal, Amolak Ratan Kalra, Sungjai Lee, Debbie Leung, Luke Schaeffer, Pulkit Sinha, Graeme Smith
Mathematical Physics
On Conservative Statistical Riemann Surfaces
Hanwen Liu
We establish a correspondence between information geometry and gauge theory. First, we define an important class of statistical
manifolds, that is normalized and satisfies a conservation field equation. Second, we prove that for a conservative statistical structure on an orientable surface, the Chebyshev 1-form is constrained to be harmonic, and the traceless part of the Amari–Chentsov tensor descends to a holomorphic cubic differential. Then, we demonstrate that normalized conservative statistical structures are geometrically generated by solutions to the scalar Tzitzéica equation on
Higgs bundles with general linear holonomy, generalizing the Labourie-Loftin correspondence. Finally, we prove that the
moduli space of normalized conservative statistical structures on a closed orientable surface of genus at least 2 is completely parameterized by a holomorphic vector bundle over the Teichmüller space, consisting of Abelian differentials and cubic differentials.
Read the full abstract ▾
Algebraic Topology
Purity of quaternionic conjugation spaces
Surojit Ghosh, Ankit Kumar and Lakshit Pande
Conjugation spaces relate the
cohomology of a space and its fixed points via a degree-halving isomorphism and admit a characterization in terms of homological purity. We extend this framework to the Klein four group, where the corresponding structures exhibit a degree-quartering behavior governed by Dickson invariants. Under a mild assumption, we prove that quaternionic conjugation spaces are homologically pure. As an application, we show that such spaces are both $\mathcal{K}_4$-maximal and $\mathcal{K}_4$-Galois maximal, establishing a connection with Smith–Thom type inequalities in real algebraic geometry.
Read the full abstract ▾
Optimization and Control · Cross-listed from eess.SY
Transfer Learning of Multiobjective Indirect Low-Thrust Trajectories Using Diffusion Models and Markov Chain Monte Carlo
Jannik Graebner and Ryne Beeson
Preliminary low-thrust spacecraft mission design is a global search problem characterized by a complex solution landscape, multiple objectives, and numerous local minima. During this phase, mission parameters are often not yet fully defined, requiring new solutions to be generated at a high cadence across varying parameter values. When combined with the indirect approach to optimal control,
diffusion models can accelerate this search by learning distributions that represent high-quality initial costates. However, generating training data remains expensive, and opportunities exist to better exploit past data. We propose a transfer-learning framework that combines homotopy in a mission parameter with Markov chain Monte Carlo (MCMC) to generate training data more efficiently. The approach reformulates a multiobjective optimization problem as sampling from an unnormalized target distribution in costate space. We compare three MCMC
algorithms on a planar multi-revolution transfer in the circular restricted three-body problem, with homotopy in the system mass parameter. The results show that gradient-based MCMC variants achieve the best trade-off between sample quality and computational cost. For the test transfer, the proposed framework generates 40 % more feasible solutions and achieves a higher-quality Pareto front than a state-of-the-art indirect approach based on adjoint control transformations and gradient-based optimization. Finally, the MCMC-generated samples are used to fine-tune a
diffusion model conditioned on the mass parameter, enabling it to learn a global representation of the underlying solution distribution and efficiently generate new solutions. These findings establish the transfer-learning framework as a practical method for efficiently solving indirect trajectory optimization problems with varying parameters.
Read the full abstract ▾
Mathematical Physics
An exact spacetime polymer gas for finite-temperature $\mathbb Z_N$ homological quantum code
Nafiz Ishtiaque and Shanto Chakroborty
We study finite-temperature $P$-form $\mathbb Z_N$ homological codes via an exact finite-Trotter
quantum-to-classical map to a $(d+1)$-dimensional spacetime model with electric and
magnetic topological background charges. The resulting background-resolved partition functions admit an exact reformulation in terms of closed
magnetic and electric defect
polymers, with opposite-species interactions governed by linking phases. By bounding this complex
polymer gas by positive same-species hard-core majorant gases, we obtain an explicit low-activity criterion under which all background-dependent partition functions are uniformly controlled and homologically nontrivial
polymers are exponentially suppressed on the scale of the spacetime systole. We also derive an exact higher-form Kramers-Wannier duality exchanging electric and
magnetic backgrounds, Wilson and 't Hooft operators, and $P$-form and $(d-P)$-form theories. Finally, for
prime $N$, we identify an exact source-free gauge-theory specialization coupled to the plaquette random-cluster model, which imports sharp phase-transition results on special geometries into the spacetime framework.
Read the full abstract ▾
Information Theory · Cross-listed from cs.LG
A Communication-Theoretic Framework for LLM Agents: Cost-Aware Adaptive Reliability
Hamed Omidvar and Vahideh Akhlaghi
Agents built on large
language models (LLMs) rely on reliability techniques such as retry, majority voting, and self-consistency developed in parallel rather than within a common analytical framework. We observe that an LLM sampled at temperature $T$ is a discrete
stochastic channel $p(y \mid x)$ in the sense of Shannon's coding theory, and use this identity as the entry point for such a framework. Each of these techniques is a special case of one of six classical reliability operators: diversity combining, hybrid retransmission, iterative generator-critic decoding, rateless sampling, structured redundant verification, and difficulty-adaptive routing. Within the framework we give two closed-form results, a noise-variance threshold above which uniform averaging beats quality-weighted averaging and a contractivity criterion for generator-critic refinement, and turn each into a design rule checkable before committing compute. We further introduce a cost-aware semantic-nearest-neighbor router whose single Lagrangian knob traverses the quality-cost frontier without retraining. Across six local and cloud channel configurations, no fixed model-technique-budget choice dominates. On a 300-item hard split of MMLU, GSM8K, and HumanEval, the router occupies the empirical Pareto frontier: at matched quality its normalized cost is $\approx$56% lower than the best fixed technique, and at matched cost it improves quality by $\approx$7% (26% over single-shot decoding). Judge-free grading of 6,900 outputs confirms that the judge tracks correctness at the item level and that the router's advantage over a fixed policy persists at most cost settings; with internal
reasoning enabled, the gain of external refinement shrinks but remains significant for our incremental-redundancy operator. These results argue for consolidating reliability techniques into a single tunable layer informed by channel coding.
Read the full abstract ▾
Numerical Analysis · Cross-listed from cs.CE
Rao-Blackwellized Markov chain Monte Carlo Light Transport
Sascha Holl, Gurprit Singh and Hans-Peter Seidel
In light transport simulation, Markov chain Monte Carlo methods are particularly effective at exploring regions with complex lighting characteristics. However,
estimator variance is a central concern across Monte Carlo methods in general. In light transport, high variance directly manifests as increased noise or, equivalently, longer rendering times at fixed image quality. Variance reduction techniques based on Rao-Blackwellization have proven particularly effective. In practice, however, the RB approach traditionally used in light transport, waste-recycling, can yield little to no measurable variance reduction, a fact we empirically confirm in this work. Motivated by this lack of effective variance reduction, we introduce a novel RB technique for the general-purpose Metropolis-Hastings
algorithm that is computationally efficient and achieves substantial variance reduction. We show that this method consistently outperforms waste-recycling in terms of both variance reduction and convergence speed. Building on this result, we adapt the proposed RB approach to the recently introduced general-purpose Jump Restore
algorithm, where it similarly achieves substantial variance reduction and accelerated convergence. Through extensive experiments in light transport simulation, we demonstrate that our \gls{rb} technique significantly outperforms the traditional approaches for both MH-based light transport
algorithms and Jump Restore Light Transport, under both equal-time and equal-sample-count comparisons.
Read the full abstract ▾