vector field analysis

This work investigates whether the drift field in single-step generative models can be interpreted as the gradient of a scalar loss function—that is, as a conservative vector field. Theoretical analysis reveals that standard position-dependent normalizations, such as those based on Gaussian kernels, generally yield non-conservative drift fields, with the Gaussian kernel itself being a benign exception. To address this, the authors propose a sharp-kernel normalization technique that restores conservativeness for any radial kernel, thereby enabling the construction of an explicit training objective. Experimental results demonstrate that training with conservative drift fields performs on par with or better than non-conservative alternatives, offering a clearer and more principled optimization target for generative modeling.

Modeling sparse vector field observations (e.g., wind, ocean currents) on two-dimensional manifolds—such as the Earth’s surface—poses challenges due to intrinsic curvature, boundary effects, and physical constraints (e.g., irrotationality or incompressibility). Method: We propose the first discrete intrinsic vector-valued Gaussian process (GP) model defined on arbitrary triangle meshes. By coupling discrete differential geometry—specifically covariant derivatives and Hodge decomposition—with intrinsic GPs, we construct a geometry-aware covariance operator that explicitly incorporates manifold curvature, boundary conditions, and physical priors. Efficient inference is achieved via variational approximation and sparse GP techniques. Results: Our model significantly outperforms Euclidean GPs and standard interpolation methods in global wind-field downscaling and sparse ocean-flow reconstruction. It accurately recovers complex circulation patterns, quantifies predictive uncertainty, and—uniquely—unifies geometric structure with physical constraints within a single probabilistic framework.

Existing learned plasma transport surrogates suffer from instability in long-term autoregressive prediction due to neglecting charge conservation, density positivity, and Poisson compatibility. This work addresses these issues by introducing FluxNet—a conservative finite-volume-based model—evaluated on a controlled one-dimensional drift-diffusion-Poisson benchmark. By incorporating structure-preserving flux corrections, positivity constraints, and Poisson field reconstruction, FluxNet enforces physical consistency throughout the simulation. Experimental results demonstrate that discrete conservation structure is critical for long-horizon rollout stability, outweighing the importance of single-step prediction accuracy. Across 64 test configurations, FluxNet achieves the lowest rollout mean squared error (MSE) in 60 cases, with a primary experiment yielding an error as low as $7.35 \times 10^{-9}$—significantly outperforming non-conservative baselines, whose errors reach the order of $10^{1}$.

This work addresses the lack of theoretical justification for acceleration mechanisms in consistency models by proposing the first rigorous convergence analysis framework. Methodologically, it constructs a multi-step consistency model and introduces backward trajectory mapping coupled with KL-divergence analysis. The main contributions are: (i) the first proof that $O(log(d/varepsilon))$ iterations suffice to achieve $O(varepsilon^2)$ KL-divergence convergence; (ii) optimality of this bound under non-smooth distribution assumptions; (iii) unified characterization of both learning error and sampling discretization error; and (iv) guaranteed optimal convergence rate even under the weakest possible data distribution assumptions, supporting constant step sizes. These results establish a solid theoretical foundation for consistency modeling and reveal the fundamental mechanism whereby generation quality improves rapidly—logarithmically—in the number of iterations.

To address the weak long-term dependency capture and low training efficiency of Neural Controlled Differential Equations (NCDEs) in modeling irregularly sampled multivariate time series, this paper proposes Log-NCDEs—a novel framework that introduces the Log-ODE method into NCDEs for the first time. Grounded in rough path theory, Log-NCDEs explicitly encodes the non-commutative structure of vector fields via Lie bracket expansions to construct high-order path signatures. This design substantially improves solution approximation accuracy and gradient stability, thereby enhancing long-horizon temporal dependency modeling. Extensive experiments on real-world and synthetic datasets with up to 50,000 irregular observations demonstrate that Log-NCDEs consistently outperforms state-of-the-art models—including NCDE, NRDE, LRU, S5, and Mamba—in both predictive accuracy and training speed. Log-NCDEs thus establishes a new paradigm for irregular time series modeling that unifies theoretical rigor with practical efficiency.

This work addresses the challenge of convergence in finite-particle systems arising from the non-conservative nature of displacement-driven velocity fields in first-order generative modeling. To overcome this, the authors propose a conservative drift formulation based on the gradient of kernel density estimation (KDE), constructing the velocity field as the smoothed difference between data and model score gradients. This approach yields, for the first time, a theoretical convergence rate for finite-particle systems in continuous time. Key contributions include elucidating the critical role of self-interaction terms, providing explicit dependence of integration constants on kernel bandwidth, and introducing a novel sharp-kernel decomposition perspective for non-conservative settings. In $\mathbb{R}^d$, the method achieves a root-mean-square velocity convergence rate of $N^{-1/(d+4)}$ or an optimized rate of $N^{-(2-\beta)/(2(d+4-\beta))}$, which directly translates into one-step generation error guarantees.

This study addresses the challenge of alarm fatigue in continuous model monitoring caused by high false positive rates of existing drift detectors, which undermines monitoring reliability. It presents the first systematic evaluation of the cumulative false positive behavior of five widely used methods—Population Stability Index (PSI), Kolmogorov–Smirnov (KS) test, Maximum Mean Discrepancy (MMD), Least-Squares Density Difference (LSDD), and adversarial validation—under continuous monitoring settings, incorporating Bonferroni correction for multiple hypothesis testing. The empirical analysis reveals that PSI exhibits markedly improved stability when sample sizes exceed 200, whereas KS, MMD, and LSDD demonstrate greater reliability with smaller batch sizes. While Bonferroni correction effectively suppresses false positives, it concurrently reduces detection sensitivity. These findings offer practical guidance for selecting batch sizes and calibrating detectors in real-world deployments, balancing robustness and responsiveness.

We study long-horizon deployment of a frozen predictor under dynamic covariate shift. A time-domain Poincaré inequality reduces temporal risk volatility to derivative energy, and a Jacobian-velocity theorem identifies directional tangent energy along the deployment path as the governing quantity under explicit along-path regularity and domination assumptions. Under low-rank drift, that quantity reduces to directional Jacobian energy in the drift subspace, motivating drift-aligned tangent regularization (DTR) and a matched monitoring proxy. Rather than smoothing the network isotropically, DTR penalizes sensitivity only along estimated drift directions. We validate the theorem-to-method pipeline in four experiments: a synthetic benchmark for the time-domain inequality, a controlled synthetic comparison against isotropic Jacobian regularization, and two frozen-deployment studies on the UCI Air Quality and Tetouan power-consumption datasets. DTR reduces risk volatility and directional gain in the controlled low-rank regime, beats isotropic smoothing there, and gives validation-selected deployment gains on both real datasets when the Air Quality drift subspace is estimated from target-orthogonal sensor motion. Moderate drift-subspace misspecification is tolerable while orthogonal misspecification largely removes the benefit.

This work addresses a systematic bias of order O(1/n) in drift models under small-batch training, induced by the self-normalizing property of the softmax function, which compromises centroid estimation accuracy. The authors propose an Analytic Bias Correction (ABC) method that explicitly models the dominant bias term using within-batch statistics and corrects the empirical centroid via a closed-form plug-in estimator. ABC is the first approach to analytically quantify and correct this bias, reducing the error to O(1/n²) without increasing first-order variance or violating convex hull containment. Experiments demonstrate that ABC significantly lowers FID and accelerates convergence on CIFAR-10, with pronounced improvements in small-batch settings; synthetic experiments further validate the theoretical bias order.