Computer Vision
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Monday 9 November
Cross-view image sequence geo-localization aims to determine the geographic location of a sequence of ground images by matching it against geo-tagged satellite imagery. Although sequential observations provide complementary scene information, substantial overlap between adjacent images introduces feature redundancy, while conventional fusion strategies may weaken spatial relationships essential for cross-view matching. Therefore, we propose a cross-view image sequence geo-localization method via Attention-guided Sequence fusion and Metric Learning (ASML). Specifically, a Sequential Content Fusion (SCF) module models cross-image feature correlations to suppress redundant content and adaptively preserve geographically informative non-overlapping regions. Next, a Spatial Feature Aggregation (SFA) module is used to capture and aggregate the spatial and semantic information of local features, so as to maintain the spatial layout of the image sequence. Finally, a Geo-InfoNCE loss function is proposed, which comprehensively considers the feature distance and geographic distance for mixed-space metric learning, thus improving the ability of the model to distinguish hard negative samples. A new cross-view image sequence dataset CrossRoad-66K is established to provide more complex scenarios. Extensive experiments and visualizations on the SeqGeo and CrossRoad-66K datasets demonstrate that compared to state-of-the-art (SOTA) methods, the proposed method ASML achieves an improvement of 3.37% and 2.10% at R@1, respectively.
General Relativity and Quantum Cosmology
Deviations from Kerr: Critical curves and photon-ring critical parameters for general observer inclinations
We study critical curves and photon-ring critical parameters in stationary, axisymmetric spacetimes with separable null geodesics and Kerr angular structure. A Kerr-relative decomposition of the radial metric functions gives analytic first-order departures of these observables for observers at general inclination. The critical curve's signed normal displacement factorizes as $\tilde u_n=\tilde F_{\mathrm K}\mathscr C$, where $\tilde F_{\mathrm K}$ depends only on the Kerr reference geometry and $\mathscr C$ is an explicit combination of optical metric deviations. The demagnification exponent, time delay, and azimuthal advance are evaluated at a common position angle on the observer screen. Intrinsic relations among the exact observables express the optical information shared by the curve, time delay, and azimuthal advance, while demagnification additionally depends on radial instability. Applications to representative metrics illustrate how different radial deformations produce distinct departures from Kerr around the critical curve. Comparisons with exact finite-deformation profiles and independent numerical calculations verify the first-order predictions on regular, accessible exterior photon-shell branches.
Read the full abstract
Quantum Physics
Parameterized equivalence checking of quantum circuit generators via weighted automata
A generator builds a quantum circuit from the number of qubits and a bit string. We check a generator against a reference for all parameter values under six equivalence relations. For ladder programs, we prove that above a computable threshold the circuit matrices form a uniform matrix product operator, a weighted automaton of dimension independent of the number of qubits. Equivalence reduces to the zeroness of a difference automaton, checked modulo primes. The answer is a counterexample or "equal", exact or with a computed error bound. Experimentally, the method found errors missed by fixed-size checks and produced no incorrect verdicts.
Read the full abstract
Artificial Intelligence
Evaluating AI Agents for Industrial-Level Chip Floorplanning: A Framework and Empirical Study
Floorplanning is an early step in turning a logical circuit into a physical chip: it positions large circuit modules and constrains the space available for subsequent logic placement and wiring. These decisions affect congestion and timing, but their physical consequences emerge through downstream implementation. For AI agents, the task requires translating engineering constraints into spatial decisions whose quality cannot be judged by successful tool execution alone. We present an evaluation framework that connects submitted layouts to a prescribed downstream implementation flow, checks geometric validity, measures physical quality, and reports resource use. On one industrial design whose 99 macros occupy approximately 71% of a fixed L-shaped layout region, we compare our specialized system Nova-SI with general-purpose engineering agents and agents restricted to file access, obtaining complete physical measurements for 39 layout outputs. Under our 0–100 composite quality score, Nova-SI achieves a mean of 57.97 without online feedback from the prescribed power-grid and placement evaluation, compared with 31.17 for the strongest measured general-purpose configuration, GPT-6 Astra. The difference is supported by raw congestion and timing measurements and persists across scoring-sensitivity checks. Layout-level analysis further shows that similar global utilization can conceal substantially different local congestion. Together, the protocol and case study assess agents through the physical consequences of their delivered layouts and provide evidence about system performance on this industrial instance.
Read the full abstract
Astrophysics of Galaxies
Dust in cold and warm dark matter in a suite of zoom-in simulations of Milky Way analogues
We use a suite of hydrodynamical zoom-in simulations of Milky Way analogues to investigate how previously identified trends in dust properties of galaxies can be utilised to infer the underlying dark matter model. We use a sample of 24 comparison simulations comparing a cold dark matter (CDM) and 3.5 keV warm dark matter (WDM) particle, primarily using a dust-to-metals (DTM) ratio to post-process the dust in the simulation, with alternative methods explored. We find a preference for higher dust masses in CDM over WDM, with a median increase of 6.6 per cent in the CDM sample at redshift $z=0$, and with a peak median difference of 25 per cent at $z\approx0.25$. Both CDM and WDM samples are highly concentrated, with approximately 90 per cent of the dust located in the inner region of the halo, though slightly higher in CDM. Contrary to previous work, we find that the Gini coefficient does not consistently show a difference in comparisons of dark matter models, with differences primarily driven by high sensitivity to the unique structure of individual haloes. Finally, we conduct an analysis of some common dust scaling relations, finding that the differences between the two samples in the dust-to-stellar mass and dust-to-SFR functions do not constitute a statistically significant difference at $z=0$ with any dust production mechanism. However, at redshifts where the difference in dust mass is greatest, the alternative dust production methods show larger separations between CDM and WDM samples in total mass and scaling relations.
Read the full abstract
Astrophysics of Galaxies
Could globular clusters in the vicinity of the galactic centre have contributed stars to the nuclear star cluster in the past?
The results of direct N-body simulations of three Milky Way bulge globular clusters - Terzan 2, Terzan 4, and Terzan 5 - evolving in a time-varying galactic potential derived from the IllustrisTNG-100 cosmological database over 8 Gyr are presented. Incorporating full stellar evolution and compact remnant formation, we quantify the probability and efficiency of stellar accretion onto the Galactic nuclear star cluster (NSC) during close pericentre passages of these three clusters. To parametrize the effect of dynamical friction in the dense central environment, we introduce a velocity reduction factor (Vred) ranging from 1.00 to 0.10. It was found that all three clusters contribute stars to the NSC even in the absence of velocity reduction, with a combined bound stellar count of 730 stars and a total accreted mass of 278 Msol at Vred = 1.00, growing to 5420 stars and 1175 Msol at Vred = 0.10. Terzan 4 is the dominant contributor across all velocity reduction factors, reflecting its more frequent and closer pericentre passages near the Galactic centre. The mass distribution of accreted stars reveals a bimodal structure: a low-mass component below 1 Msol whose capture is independent of Vred, and a prominent high-mass tail extending to 20 Msol corresponding to neutron stars and black holes, which grows with decreasing Vred. Notably, Terzan 4 is the only cluster for which white dwarfs are also captured by the NSC at every Vred value. Our results suggest that the Terzan family represents a more significant source of compact remnants for the Galactic NSC than previously studied clusters, while confirming that tidal stripping of surviving globular clusters alone cannot account for the total NSC mass.
Read the full abstract
Fluid Dynamics
Variational principles for dissipative continua: transport theorems, inertia, and the minimum-maximum duality
Variational principles for dissipative processes come in two families: minimum principles at nonequilibrium stationary states (Helmholtz, Kirchhoff, Prigogine) and maximum principles for relaxation in continuum mechanics (Ziegler, Hackl, Fischer). We contribute to the understanding of the thermodynamic extremal principle (TEP) by investigating its extremal properties, and uncover the connection between the two families, whose distinction was recently made precise for diffusion: the extremal character is decided by the regime, which is either relaxation at frozen state, as in a closed system (maximum dissipation), or a driven stationary state in an open system (minimum dissipation), while the underlying Rayleighian functional is convex and minimized in both cases. We further examine the Reynolds transport theorem, a useful tool for fixing the energy rates of continua, and the role of inertia, recently argued to lie beyond the reach of Rayleighian principles. With inertia the driving force must be held fixed during the variation, which is less elegant than in the inertialess case, where it is a state function. Writing the energy rate as rates times driving forces still yields the correct equations, and an extremal statement survives whenever the frozen force does not depend on the varied rate. This holds directly for unsteady Stokes flow, and for Navier-Stokes flow after a change to material coordinates, in which the particle acceleration is the second time derivative of the state and the convective term, nonvariational in the Eulerian description, does not appear. The framework is tested on the Navier-Stokes family (Stokes, unsteady Stokes, and Navier-Stokes flow, in order of increasing complexity) and on dissipative Lagrangian mechanics.
Read the full abstract
Image and Video Processing
Leveraging High-Activity Post-Therapeutic I-131 Scans for Deep Learning Reconstruction of Extremely Sparse Planar Images in Differentiated Thyroid Cancer
Background: Diagnostic I-131 whole-body planar imaging in differentiated thyroid cancer (DTC) is count-limited by low administered activity, and quality worsens under rapid acquisition. Models trained on such data struggle because their targets lack photon statistics. Purpose: To test whether high-quality post-therapeutic (PT) I-131 scans can serve as superior training references for enhancing sparse diagnostic planar images, using transfer learning and input-image addition. Methods: In this single-center prospective study, 111 patients were enrolled in PT training (n = 19), diagnostic follow-up (FU) training (n = 58), and FU paired test (n = 34) cohorts. Five Restormer-based models were trained: FU_DLx2 on FU labels; PT_DLx2 transfer-learned on PT data; PT_DLx1+ and PT_DLx2+ with input-image addition; PT_DLx6 as a PT-only reference. Test-time processing used two-step denormalization and Poisson re-scaling to preserve noise statistics. Quality was assessed by PSNR, SSIM, and a blinded dual-reader evaluation of all 34 test-cohort patients. Results: PT_DLx2 achieved the highest PSNR (34.4 +/- 3.6 dB) and SSIM (0.81 +/- 0.07). PT_DLx1+ and PT_DLx2+ were comparable (33.8 and 33.3 dB; SSIM 0.80). FU_DLx2 ranked below the unprocessed fast scan (30.3 vs. 32.0 dB; 0.76 vs. 0.79). Subjectively, PT_DLx2+ scored highest in image quality (4.54 +/- 0.60) and diagnostic confidence (4.54 +/- 0.60), surpassing the reference standard (4.19 +/- 0.83 and 4.31 +/- 0.81; P = 0.005 and P = 0.008). FU_DLx2 and PT_DLx6 scored lowest. Conclusion: Count-rich PT scans provide effective training priors for reconstructing extremely sparse images. Input-image addition and cross-domain transfer learning from PT data outperformed direct training on count-limited diagnostic data, supporting 2x scan acceleration with quality matching or exceeding the unaccelerated standard scan for selected PT-derived configurations.
Read the full abstract
K-Theory and Homology
$\mathrm{RO}(G)$-graded norms for de Rham-Witt forms
We make two case studies of $\mathrm{RO}(G)$-graded homotopy in connection with $p$-adic cohomology. First, we identify the $\mathrm{RO}(\mathbf T)$-graded $\mathrm{TC}^-$ and $\mathrm{TP}$ of quasiregular semiperfectoid rings as multi-Rees algebras for the Nygaard and $\phi^\bullet(I)$-adic filtrations. We additionally compute $E^{h\mathbf T}_\bigstar$ and $E^{t\mathbf T}_\bigstar$ when $E$ is a general even $\mathbf E_\infty$-ring with $\mathbf T$-action, and compute $(\mathrm{TR}^n)^{h\mathbf T/C_{p^n}}_\bigstar$ in terms of iterated Nygaard filtrations. Next, we compute the $\mathrm{RO}(\mathbf Z_p)^\land$-graded $\mathrm{TR}$ of smooth algebras over a perfectoid base, providing a pushout of the works of Hesselholt and Angeltveit-Gerhardt along the base case $\mathbf F_p$. Using this, we extend Angeltveit's Norm map of Witt vectors to the de Rham-Witt complex, which participates in a "Witt coordinates" decomposition of the $p$-primary Adams operations.
Read the full abstract
Computer Vision
Four-Step Text-Bearing Image Generation via Distillation with Local Response Regularization and Teacher-Guided Training
Generating text-bearing images requires coherent scenes and accurate characters. Large text-to-image models improve both capabilities through repeated denoising, at substantial sampling cost. Few-step distillation reduces this cost but requires the student to organize scenes and refine glyphs within a short trajectory. We present a four-step distillation framework combining local response regularization with teacher-guided text training. Built on distribution matching, the framework distills the GLM-Image DiT while retaining its autoregressive visual prior, glyph encoder, and VAE. A path-length penalty promotes consistent local responses of the conditional denoising field. Progressive training combines synthetic strings with teacher-generated imagery to preserve character accuracy and scene-integrated typography. The student achieves GenEval Overall 0.84 and CVTG-2K word accuracy 0.5045, compared with 0.81 and 0.1690 for four-step TDM. A coefficient study favors $10^{-5}$, and Qwen-Image experiments show additional gains from response regularization and text supervision. Single-image inference takes 1.742 s, compared with 5.818 s for the teacher, yielding a $3.34\times$ speedup at the same 40 GB GPU memory requirement.
Read the full abstract
Computation and Language
SMO: Compiling Natural Language Skills into Graphs of Reusable State-Move-Outcome Bricks
Large language model agents increasingly rely on reusable natural-language skills, yet growing skill collections remain accumulations of isolated text artifacts that are difficult to compare, deduplicate, and compose. We argue that the reusable content of a skill lies not in its surface wording but in its procedural transitions: which operation to execute under a given state and which observable outcome to produce. Based on this observation, we propose State-Move-Outcome (SMO), a compilation-like framework that decomposes raw skills into reusable procedural bricks and organizes admitted bricks as an externalized symbolic structure. When compiling a skill, SMO first retrieves and composes existing bricks, dynamically connects them through Outcome-to-State conditions, and creates a candidate brick only for an uncovered procedure; the candidate becomes reusable only after validation and admission. Experiments on SkillsBench v1.1 demonstrate that SMO represents 232 rawSkills from 87 underlying tasks with 95 shared bricks, reducing the corresponding skill representations from 366,472 to 149,154 tokens–a $\mathbf{59.3}\times$ reduction and $\mathbf{2.46}\times$ compression. Across 232 skill-conditioned evaluation instances, SMO raises the average task score from $\mathbf{0.6379}$ to $\mathbf{0.7478}$ and increases 5/5 Pass instances from $\mathbf{138}$ to $\mathbf{164}$, while matching or outperforming rawSkill on $\mathbf{208}$ of $\mathbf{232}$ instances. It also reduces average runtime token consumption by $\mathbf{38.38\%}$. These results show that compiling natural-language skills into shared procedural units can jointly improve skill reuse, representation efficiency, and downstream execution.
Read the full abstract
Computer Vision
Leveraging Complementary Event Frame and Voxel Representations for RGB-Guided Multimodal Pre-training
In this paper, we propose CM3AE 2.0, a unified RGB Frame and Event-Voxel/-Frame pre-training framework. By jointly learning multimodal spatiotemporal representations and exploiting latent cross-modal dependencies through cross-modal generation, the framework enhances multimodal information interaction and collaborative modeling capabilities. Specifically, we first propose an Event-aware Spatiotemporal Masked Reconstruction (ESMR) strategy to learn unified multimodal spatiotemporal representations. It leverages event responses to guide masking toward dynamically salient regions. Through spatiotemporal masked reconstruction, it learns spatial structures and their temporal variations from continuous RGB-Event observations. In addition, it incorporates multimodal fusion reconstruction and cross-modal contrastive learning to facilitate the modeling of cross-modal complementary information and the learning of shared representations. Building upon this, we further propose a Hierarchical Cross-modal Generation (HCG) strategy to learn latent cross-modal dependencies through visual and language generation. At the visual level, Event-conditioned RGB generation is performed using a pre-trained diffusion model to capture visual associations between the two modalities. At the language level, video-level text description generation based on joint RGB-Event representations establishes connections between multimodal visual features and linguistic semantics, further enriching the learning of cross-modal dependencies. Extensive experiments on four diverse downstream tasks demonstrate that CM3AE 2.0 consistently improves performance, validating the effectiveness and generalization capability of the proposed pre-training framework.
Read the full abstract
Logic
Adding minimal fresh subsets to $\aleph_\omega$
Assuming the existence of a measurable cardinal, we produce generic extensions $V_1\subsetneq V_2$ of the universe with the same cardinals, cofinalities, and bounded subsets of $\aleph_\omega$, such that $V_2$ is generated over $V_1$ by a cofinal $\omega$-sequence in $\aleph_\omega$, and $V_2$ is a minimal generic extension over $V_1$, i.e., every set of ordinals in $V_2\setminus V_1$ generates $V_2$ over $V_1$.
Read the full abstract
Robotics
DPGS-SLAM: Diffusion-Prior and Provenance-Aware Monocular Gaussian Splatting SLAM for Dynamic Visual Sensing
This paper presents DPGS-SLAM (Diffusion-Prior and Provenance-Aware Monocular Gaussian Splatting SLAM) for dense mapping in environments containing dynamic objects. Objects in dynamic motion remain a significant challenge for dense visual SLAM since dynamic observations can contaminate camera tracking and scene reconstruction. Although 3DGS (3D Gaussian splatting) enables efficient dense mapping, existing 3DGS-SLAM systems may exhibit instability when dynamic objects violate their underlying static-scene assumption. Excluding dynamic observations can reduce their influence on tracking and mapping, but may leave occluded static backgrounds missing or weakly constrained. Our proposed DPGS-SLAM is a monocular RGB SLAM framework that supports provisional representations of insufficiently observed static backgrounds through asynchronous diffusion-based background completion and provenance-aware Gaussian mapping. Dynamic regions are identified through optical-flow-based motion discovery and temporally consistent mask propagation by SAM2. For selected keyframes, a fine-tuned diffusion inpainting model generates plausible background appearance in occluded regions, while metric depth predicted from completed images initializes provisional Gaussian primitives. Provenance labels distinguish generated primitives from observation-supported primitives, excluding generated content from camera-pose estimation and observed-map optimization. When previously occluded backgrounds become visible, eligible generated primitives are removed, and observation-supported primitives are independently initialized from real RGB observations and predicted depth. Experimental results demonstrated competitive camera tracking, with average ATE RMSE values of 1.2 cm on TUM RGB-D and 2.7 cm on BONN RGB-D datasets. On StaDy4D, DPGS-SLAM improved average PSNR by 4.60 dB over WildGS-SLAM using only monocular RGB input.
Read the full abstract
Computer Vision
SAFE-Hair: Scalp-Anchored Fields for Exportable Single-View Hair Reconstruction
Single-view 3D hair reconstruction has progressed rapidly, but many methods optimize for visual plausibility rather than producing strand assets that can be directly used in grooming, editing, or simulation pipelines. We propose SAFE-Hair, a framework that reconstructs explicit strand grooms from a single portrait by decoding hair from persistent follicles in a canonical scalp-UV domain. Given a fitted template head, SAFE-Hair projects self-supervised image features onto the scalp, uses a conditional Rectified Flow model to generate latent scalp fields, and decodes occupancy, segment directions, length allocation, and total length at a fixed set of scalp follicles. This representation enforces scalp-rooted strand identities by construction and enables direct export with follicle IDs and guide-family metadata. We further introduce attachment, collision, and bending priors to improve geometric validity without per-subject test-time optimization. To evaluate both reconstruction fidelity and asset usability, we introduce HairBench-3D, a benchmark built from public 3D hair assets with standardized alignment, rendering, resampling, and metrics for geometry, direction consistency, invisible-region completion, silhouette agreement, and collision. SAFE-Hair improves Chamfer distance from 3.74 mm to 3.28 mm, F-score from 77.9 to 81.7, invisible-region F-score from 66.1 to 69.8, and penetration ratio from 4.8% to 3.9% over the strongest baseline.
Read the full abstract
Applied Physics
Split-Ring Couplers for On-Skin Wi-Fi Communication via an SSPP Waveguide Textile
Textile spoof surface plasmon polariton (SSPP) waveguides offer guided on-body links, but their body-facing ground conductors obstruct coupling from on-skin wireless nodes. This letter presents a split-ring coupling interface using local ground recesses at the textile terminations, without patterned garment-side launchers. The capacitively loaded rings are supported by polydimethylsiloxane carriers with a thickness of 2~mm. An image-current model and full-wave simulations examine the effect of backing-ground spacing and determine the recess geometry. For a 450~mm textile waveguide on a water-filled body phantom, the measured $S_{21}$ at 2.45 GHz was $-$18.8~dB, versus $-$60.9~dB without the textile. A 625-position scan showed a maximum 5.1 dB reduction relative to the nominal position within the measured joint $\pm$5~mm placement region. A separate two-node IEEE~802.11n test demonstrated a 65~Mb/s physical-layer (PHY) rate with zero packet errors among 300 frames at $-$20~dBm transmit power through the coupler link. These results establish high-rate, placement-tolerant on-skin Wi-Fi communication through an SSPP waveguide textile.
Read the full abstract
Analysis of PDEs
Uniqueness of positive solutions of mixed Laplacian and fractional Laplacian in $\mathbb R$
We establish the uniqueness and nondegeneracy for positive finite-energy solutions of the following one-dimensional mixed Schrödinger equation \begin{equation} -Q” + (-\Delta)^s Q + Q = |Q|^p Q \quad \text{in } \mathbb R, \end{equation} where $0<s<1$ and $p>0$. For every $0<s<1, p>0$, we obtain the corresponding results for positive solutions whose linearized operators have Morse index equal to one. We extend the nodal analysis developed by Frank-Lenzmann-Silvestre to the mixed operator establishing the monotonicity of the quotient of a hypothetical nontrivial even kernel element by $-Q'$. We then combine this monotonicity with oscillation estimates to prove nondegeneracy, and uniqueness follows by continuation. Moreover, for every $1/2\le s<1,p>0$, both uniqueness and nondegeneracy hold for positive solutions. Our novelty is to use the properties of Green kernel of the mixed operator to derive the upper bound and lower bound on the maximum of each solution. This peak estimate allows us to prove uniqueness by direct comparison and to establish nondegeneracy.
Read the full abstract
Data Structures and Algorithms
The Power of Migration on Identical Machines is Two
We study the power of migration in preemptive scheduling on identical parallel machines with time windows. If an instance admits a migratory schedule on $m$ machines, how many machines are needed to schedule it without migration? Kalyanasundaram and Pruhs (2001) showed that $6m-5$ machines suffice and gave a polynomial-time construction using $12m-5$ machines, whereas $2m-1$ machines may be necessary. We close this gap, which has been open for more than 25 years: $2m-1$ machines always suffice, and a corresponding non-migratory schedule can be computed in strongly polynomial time. Our proof has two main technical ingredients. First, we introduce supercanonical schedules and prove that the conflict graph of every such schedule on $m$ machines is $(2m-1)$-colorable. The key property is that whenever a job pauses, the jobs running alongside it just before and just after the pause separate the jobs processed entirely during the pause from the rest of the conflict graph. We use these separators and Kempe exchanges to compute a coloring in polynomial time. Second, we construct a supercanonical schedule with polynomially many processing blocks by shrinking job windows as far as possible while preserving feasibility and solving a min-cost flow problem.
Read the full abstract
Physics Education
Aircraft speed and height from Doppler-shifted sound and flight-track data
The changing pitch of a passing aircraft is a familiar example of the Doppler effect. We use audio recordings of 40 aircraft passages to estimate the speed and height by two methods based on the Doppler effect. The endpoint method is an analytical model that uses the limiting frequencies at the start and end of each recording and the slope near the middle of the frequency-time curve. The second method fits the whole curve. Results from both methods are compared with independent flight-tracking data. The endpoint method consistently underestimates speed and height because the recorded frequencies have not approached their limiting values. The full curve fit agrees with the tracking data within about 3% in speed and 12% in height for steady passages, with poorer agreement when speed or height changes. Software developed with AI assistance made it practical to analyze this many passages. The comparison offers undergraduate students an example of testing model assumptions and identifying systematic errors in real measurements.
Read the full abstract
Superconductivity
Magnetotropic signatures of Fulde-Ferrell Pairing in a Pauli-limited superconductor
Resonant torsion magnetometry measures the angular curvature of the free energy and is sensitive to magnetic anisotropy. We calculate its response for a clean Pauli-limited superconductor with single-plane-wave Fulde-Ferrell pairing. Field rotation drives a first-order transition from the uniform condensate into the finite-momentum phase and a continuous transition into the normal state. The first-order line produces a sharp softening whose integrated weight is fixed by the magnetization jump, so the dip alone does not identify finite momentum. The distinguishing response is a broad softening inside the Fulde-Ferrell phase, caused by relaxation of the gap and pairing momentum, followed by a finite step at the upper critical field.
Read the full abstract
Machine Learning
Scale-Invariant Empirical-Bayes Laplace Approximation for ReLU Networks
ReLU networks are well-known to have continuous scale symmetries; infinitely many reparameterized weight vectors that represent the same function. Dinh et al. (2017) have shown that, by reparameterizing, the geometry of the loss landscape (e.g., curvature) of these networks can be changed arbitrarily without affecting the underlying function. Thus, any quantity that depends on the network's geometry should be defined with care. In this paper, we show that the Empirical Bayes (EB) Laplace's approximation is not well-defined for ReLU networks, mainly due to Hessian dependence in computing the marginal likelihood, which is not invariant under ReLU scale symmetries. Thus, we can get different prior precisions, posterior covariances, and predictive uncertainties for a network having identical predictions under ReLU scaling. We define family of prior precision fields indexed by $\gamma$ and introduce $\gamma$-canonicalization, which first converts the network to its canonical representation by balancing the ratio of incoming to outgoing weights, and then fits the Empirical Bayes (EB) Laplace to the canonical representation. This procedure is equivalent to transporting the isotropic Gaussian prior from the canonical representation to the original coordinates, by producing a ReLU-scale-compatible precision field. This makes the marginal likelihood invariant under ReLU scaling, while allowing $\gamma$ to be selected by maximizing the marginal likelihood. The resulting EB Laplace approximation is well-defined on ReLU scale-equivalence classes and yields consistent prior precision selection and predictive uncertainties across functionally identical parameterizations.
Read the full abstract
Combinatorics
Spectral conditions for the existence of $\{K_2,\{C_k\}\}$-factor in $t$-tough graphs
For $t>0$, a graph $G$ is called $t$-tough if $|S|\geq t\,c(G-S)$ for every $S\subseteq V(G)$ with $c(G-S)\geq2$, where $c(G-S)$ denotes the number of components of $G-S$. Toughness is closely related to the existence of Hamiltonian cycles and factors in graphs. Chvátal [Discrete Math. 5 (1973) 215–228] conjectured that there exists a constant $t_0>0$ such that every $t_0$-tough graph of order at least three is Hamiltonian. Bauer, Broersma and Veldman [Discrete Appl. Math. 99 (2000) 317–321] constructed $(9/4-\varepsilon)$-tough graphs without a Hamiltonian path for arbitrarily small $\varepsilon>0$. For the weaker requirement of a $2$-factor, Enomoto et al. [J. Graph Theory 9 (1985) 87–95] proved that every $2$-tough graph of order at least three has a $2$-factor and that the toughness threshold $2$ is best possible. A graph $G$ is said to have a $\{K_2,\{C_k:k\geq3\}\}$-factor if and only if $G$ contains a spanning subgraph consisting of components that are either cycles or single edges. A $\{K_2,\{C_k:k\geq3\}\}$-factor can be viewed as a combination of a $1$-factor and a $2$-factor, motivating the study of such factors in tough graphs. In this paper, we establish sharp sufficient conditions in terms of the adjacency spectral radius, the distance spectral radius, and the distance signless Laplacian spectral radius for a $t$-tough graph to have a $\{K_2,\{C_k:k\geq3\}\}$-factor.
Read the full abstract
Differential Geometry
Quasipositive and almost positive sectional curvature on finite quotients of Wallach spaces
We construct nonnegatively curved metrics on the cyclic and full permutation quotients of the Wallach flag manifolds in dimensions $6$, $12$, and $24$. The metrics are quasipositively curved in all three dimensions and almost positively curved in dimensions $12$ and $24$. The cyclic quotients are orientable with fundamental group $C_3$, and their further full permutation quotients have fundamental group $S_3$; none admits any metric of everywhere positive sectional curvature. For the almost-positive metrics we describe every zero-curvature plane and prove that the zero-point loci have Hausdorff dimensions exactly $11$ and $23$, respectively. A one-step deformation has, in each dimension, an explicit zero locus with nonempty interior. Further nested deformations reduce it to a measure-zero set while leaving a common codimension-one stratum. The argument combines normal forms for horizontal commuting planes with quaternionic rank conditions and a dimension estimate for commuting compressions of skew-symmetric operators. We also determine the possible degrees of arbitrary free finite quotients of the three flag covers and derive order bounds for fundamental groups when the Euler characteristic is nonzero. We have used ChatGPT-6-astra for mathematical exploration and verified and rewritten all the results.
Read the full abstract
Fluid Dynamics
Parametric spatio-temporal reduced-order modelling via Spectral Proper Orthogonal Decomposition and Gaussian Process Regression
We present SPOD-GPR, a non-intrusive, parametric, spatio-temporal reduced-order model (ROM) that combines Proper Orthogonal Decomposition (POD), Spectral POD (SPOD), and Gaussian Process Regression (GPR). A parametric POD first projects the high-dimensional state onto a low-dimensional subspace. The reduced trajectories are then divided into windowed sections, transformed into the frequency domain, and truncated to the most energetic frequencies. SPOD is then performed independently at each retained frequency, yielding temporally coherent modes and coefficients. The parametric dependence of the POD and SPOD bases is modelled by GPR on a local chart of the Grassmann manifold after Procrustes alignment. The SPOD modal coefficients are learned both as explicit functions of the parameters and through a parameter-conditioned autoregressive model that allows for temporal extrapolation beyond the training window. Gaussian processes are sampled jointly and pathwise via Matheron conditioning, ensuring that each predictive realization carries uncertainty quantification consistent with the posterior distribution. The method is tested on four parametric examples where the parameter values and the prediction horizon lie outside the training data: the two-dimensional anisotropic Kuramoto–Sivashinsky equation in a time-periodic and a travelling-wave regime, each parametrised by one of its two coefficients; flow past a cylinder parametrised by the Reynolds number; and flow past an elliptic obstacle parametrised by its semi-axes. In all cases, SPOD-GPR reduces the normalised mean absolute error compared to a POD-GPR baseline, and improves the calibration of the predictive uncertainty, especially in the fluid flow cases, where POD-GPR is severely overconfident. This opens the way to spectral, uncertainty-aware reduced-order models for many-query and real-time applications, such as digital twins.
Read the full abstract
Computational Engineering
An open source Python code for three-dimensional topology optimization in single precision on a consumer GPU
Three-dimensional topology optimization solves linear systems with millions of unknowns at every design iteration. Consumer graphics cards run far faster in single than in double precision, and open codes for them mostly address minimum compliance. This educational article presents TOF3D, an open source Python code for density-based topology optimization in single precision on one consumer GPU. It solves four classes of problems: stress constrained minimum compliance, robust minimum compliance against manufacturing errors, compliant mechanisms and bone-like porous infill. The stiffness operator is applied without assembly, and the conjugate gradient method is preconditioned by geometric multigrid. Rounding the element matrix to single precision gave rigid body motions a spurious energy and changed compliances by up to 5.4%. Forming the element products from differences of nodal displacements reduced this error to $1.3\times 10^{-5}$ at no cost in time. Adjoint sensitivities agreed with finite differences to within $1.7\times 10^{-6}$, and the solutions agreed with Ansys Mechanical APDL to within $4.1\times 10^{-10}$ with the same element. On an RTX 3060 with 12 GB, one design iteration of a cantilever took 0.46 s with 0.5 million elements and 15 s with 33.5 million. On the same card and problem, an open GPU code that applies its operator in double precision took 3.7 and 21 s with 0.5 and 4.2 million elements and ran out of memory on larger meshes. Eight cases fixed in a protocol before the runs took 9 to 35 minutes each with 4.2 to 5.2 million elements.
Read the full abstract
Machine Learning
MambaPD: Bidirectional Selective State Space Model for Parkinson's Disease Detection from Resting-State EEG
Typical diagnosis of Parkinson's disease (PD) occurs only after significant dopaminergic loss, while resting-state EEG abnormalities like alpha slowing and reduced coherence occur earlier. While many published EEG-based PD detectors claim accuracy greater than 95%, many use epoch-level splits which leak subject identity. I introduce MambaPD, a bidirectional selective state space model for resting-state EEG-based PD detection, blending the channel mixing, patch tokenization and bidirectional Mamba encoder. It is compared with four baselines (TCN, BiLSTM with attention, EEGNet and a patch Transformer) in two public datasets (UNM, SanDiego) using subject-independent 5-fold cross-validation (checkpoints were selected on the held-out fold, so absolute estimates are optimistic). On UNM the accuracy is between 0.67 and 0.71, and MambaPD (0.714) is not statistically distinguishable from the Transformer (0.715), and TCN (0.708), with the highest sensitivity (0.82), while the lowest specificity (0.52). For the smaller SanDiego cohort, accuracy is in the range of 0.60 to 0.70, with the highest scores (0.703) from TCN and the lowest (0.601) from EEGNet, while MambaPD scores 0.631, with the lowest specificity (0.46) among the five models. The differences among the models on UNM are minor compared to the differences among folds. I report the whole pipeline, per-fold results, and the shortcomings of the protocol.
Read the full abstract
Computer Vision
Single Image Super-Resolution via Semantic Prompt and Cross-Modal Training on Content Area
Enhancing image quality through super-resolution is crucial in image and video processing. While contemporary deep learning models, such as CNNs and Transformers, excel in pixel-level enhancements, they often overlook the rich semantic context within images. To address this, we introduce a novel approach that synergizes pixel intensity with semantic information for superior detail recovery. Our method involves extracting semantic data, including textual descriptors and spatial positioning, and encoding this information. We also present a Cross-Modal Feature Fusion module for generating semantic feature maps via multimodal training, and a Prompt Traction Head to augment the backbone network with semantic insights, thereby refining image details. Our empirical results across various datasets validate the superiority of our approach.
Read the full abstract
Machine Learning
FocusBranch: Combinatorial Branch-and-Bound for $\ell_0$ Neural Network Robustness Verification
Verifying the local robustness of neural networks under $\ell_0$ perturbations is inherently combinatorial, as it must reason over all sparse subsets of $\epsilon$ perturbed pixels. This paper presents FocusBranch, a combinatorial branch-and-bound framework for $\ell_0$ robustness verification. The key insight is to make the $\ell_0$-ball decomposable and branchable using a node abstraction, which consists of a focus pixel set $\mathcal{S}$ and a level $l$, permitting at most $\epsilon-l$ pixels outside $\mathcal{S}$. This abstraction unifies two branching mechanisms: concrete nodes branch through upper shadows to fix additional perturbed locations, while super nodes branch through covering designs to cover many perturbation cases with fewer child nodes. FocusBranch proves node safety by concretizing linear relaxations over node-constrained $\ell_0$-balls and recursively branches on unknown nodes. Rather than relying on a fixed verification plan, it samples nodes across levels to estimate safe rates and costs, then selects a starting level and branching strategy predicted to minimize verification time. Experiments on fully connected and convolutional classifiers for MNIST, Fashion-MNIST, and CIFAR-10 show that FocusBranch verifies $\ell_0$ robustness effectively across architectures and outperforms state-of-the-art verifiers. The code is available at: https://github.com/simpl-group/FocusBranch.
Read the full abstract
Computer Vision
TIRAD: Text-guided Image Residual Learning for Generalist Anomaly Detection
Recent anomaly detection studies have shown that residual features, which model the discrepancy between normal and abnormal patterns, provide strong generalization to anomalies unseen during training. However, existing residual-based approaches mainly rely on visual discrepancies derived from image representations, while semantic discrepancies that can be explicitly expressed through language remain largely underexplored. To address this problem, we propose TIRAD, Text-guided Image Residual Learning for Generalist Anomaly Detection, a multimodal residual learning framework that defines the discrepancy between textual representations of normal and abnormal states as a Text Residual Feature and uses it to guide the Image Residual Feature. Specifically, a Text Fusion Module converts patch-level text residuals into a gate that modulates the image residual of each patch. The gate retains the visual discrepancies that the text residual identifies as defect-related and attenuates the others, leading to more discriminative anomaly representations. Extensive experiments on industrial benchmarks and industrial-to-medical transfer demonstrate that the proposed method outperforms existing image-centric residual learning methods.
Read the full abstract
Numerical Analysis
A Finite-Lag Markov Chain Model for Coarse-Graining Linear SDEs
In this paper, we analyze a finite-lag Markov chain approximation for coarse-grained linear stochastic differential equations. The chain uses the limiting conditional transition law of the resolved variables to advance on a prescribed time grid. Its stability and statistical properties are studied, with particular attention to the role of the lag time. The chain preserves the exact equilibrium distribution and matches the autocovariance at the chosen lag, while errors at subsequent times are quantified. For a slow-fast system, the analysis relates lag selection to scale separation and dynamical accuracy. A continuous embedding of the chain enables a unified comparison with some existing continuous-time Markovian effective models based on the temporal statistics they preserve. Numerical examples illustrate the theoretical results.