Roy Y. He

Roy Y. He, Hao Liu

We propose a novel model for decomposing grayscale images into three distinct components: the structural part, representing sharp boundaries and regions with strong light-to-dark transitions; the smooth part, capturing soft shadows and shades; and the oscillatory part, characterizing textures and noise. To capture the homogeneous structures, we introduce a combination of $L^0$-gradient and curvature regularization on level lines. This new regularization term enforces strong sparsity on the image gradient while reducing the undesirable staircase effects as well as preserving the geometry of contours. For the smoothly varying component, we utilize the $L^2$-norm of the Laplacian that favors isotropic smoothness. To capture the oscillation, we use the inverse Sobolev seminorm. To solve the associated minimization problem, we design an efficient operator-splitting algorithm. Our algorithm effectively addresses the challenging non-convex non-smooth problem by separating it into sub-problems. Each sub-problem can be solved either directly using closed-form solutions or efficiently using the Fast Fourier Transform (FFT). We provide systematic experiments, including ablation and comparison studies, to analyze our model's behaviors and demonstrate its effectiveness as well as efficiency.

Roy Y. He, Sung Ha Kang, Jean-Michel Morel

Image vectorization converts raster images into vector graphics composed of regions separated by curves. Typical vectorization methods first define the regions by grouping similar colored regions via color quantization, then approximate their boundaries by Bezier curves. In that way, the raster input is converted into an SVG format parameterizing the regions' colors and the Bezier control points. This compact representation has many graphical applications thanks to its universality and resolution-independence. In this paper, we remark that image vectorization is nothing but an image segmentation, and that it can be built by fine to coarse region merging. Our analysis of the problem leads us to propose a vectorization method alternating region merging and curve smoothing. We formalize the method by alternate operations on the dual and primal graph induced from any domain partition. In that way, we address a limitation of current vectorization methods, which separate the update of regional information from curve approximation. We formalize region merging methods by associating them with various gain functionals, including the classic Beaulieu-Goldberg and Mumford-Shah functionals. More generally, we introduce and compare region merging criteria involving region number, scale, area, and internal standard deviation. We also show that the curve smoothing, implicit in all vectorization methods, can be performed by the shape-preserving affine scale space. We extend this flow to a network of curves and give a sufficient condition for the topological preservation of the segmentation. The general vectorization method that follows from this analysis shows explainable behaviors, explicitly controlled by a few intuitive parameters. It is experimentally compared to state-of-the-art software and proved to have comparable or superior fidelity and cost efficiency.

Roy Y. He, Sung Ha Kang

We explore identifying partial differential equations (PDEs) from noisy observations of single time-space trajectories. Recent developments show the benefits of identifying PDEs in their weak forms. We investigate the use of differential Strong-form dictionaries for PDE IDENTification (S-IDENT), which enables finding more general linear and nonlinear PDEs. Building on an extensive exploration of integral-type denoised differentiation approaches, we propose to use Savitzky--Golay (SG) differentiation with an adaptive window length chosen based on Stein's Unbiased Risk Estimate (SURE). This offers a guaranteed order of accuracy while producing estimators with minimal variance. The identification process is further refined and stabilized through trimming and reduction-in-residual model selection. Numerical evidence shows that S-IDENT can successfully identify nonlinear PDEs at higher levels of noise than existing strong-form methods, while also yielding results comparable to weak-form approaches. We further verify the effectiveness of S-IDENT through comparisons with various strategies to approximate differential features. We provide numerical evidence that general differential-form dictionaries are larger and more ill-conditioned than those used for weak-form identification, yet S-IDENT does not significantly suffer from this combinatorial increase in dictionary size.

Jin Guo, Roy Y. He, Jean-Michel Morel

In 2020 Domingos introduced an interpolation formula valid for "every model trained by gradient descent". He concluded that such models behave approximately as kernel machines. In this work, we extend the Domingos formula to stochastic training. We introduce a stochastic gradient kernel that extends the deterministic version via a continuous-time diffusion approximation. We prove stochastic Domingos theorems and show that the expected network output admits a kernel-machine representation with optimizer-specific weighting. It reveals that training samples contribute through loss-dependent weights and gradient alignment along the training trajectory. We then link the generalization error to the null space of the integral operator induced by the stochastic gradient kernel. The same path-kernel viewpoint provides a unified interpretation of diffusion models and GANs: diffusion induces stage-wise, noise-localized corrections, whereas GANs induce distribution-guided corrections shaped by discriminator geometry. We visualize the evolution of implicit kernels during optimization and quantify out-of-distribution behaviors through a series of numerical experiments. Our results support a feature-space memory view of learning: training stores data-dependent information in an evolving tangent feature geometry, and predictions at test time arise from kernel-weighted retrieval and aggregation of these stored features, with generalization governed by alignment between test points and the learned feature memory.

Jin Guo, Roy Y. He, Jean-Michel Morel

Understanding how training data shape neural network predictions is a central problem in modern learning theory. In 2020, Pedro Domingos proposed an interpolation formula valid for every model learned by deterministic gradient descent. It expresses the model's prediction as an integral, along the optimization path, of a data-dependent kernel that aligns the model's gradients at the test and training data. Such a first-order characterization remains valid for models trained with batch-based stochastic optimization. In this paper, we develop second-order forms of these interpolation formulas. We show that the leading path-kernel interpolation is supplemented by a curvature-weighted interpolation term. For stochastic gradient descent, an additional sampling-induced component appears, coupling the curvature of the prediction with the covariance of mini-batch gradient noise. We also extend the representation to stochastic gradient descent with momentum, where the interpolation structure is preserved but with the weights modified by a memory-related factor. Moreover, we establish a concentration estimate for the terminal prediction, identifying the fluctuation scale around the expected second-order representation. Together, these results provide a refinement of the path-kernel interpretation of neural network prediction.

Diego Belzarena, Seginus Mowlavi, Aitor Artola et al.

Current OCR systems are based on deep learning models trained on large amounts of data. Although they have shown some ability to generalize to unseen data, especially in detection tasks, they can struggle with recognizing low-quality data. This is particularly evident for printed documents, where intra-domain data variability is typically low, but inter-domain data variability is high. In that context, current OCR methods do not fully exploit each document's redundancy. We propose an unsupervised method by leveraging the redundancy of character shapes within a document to correct imperfect outputs of a given OCR system and suggest better clustering. To this aim, we introduce an extended Gaussian Mixture Model (GMM) by alternating an Expectation-Maximization (EM) algorithm with an intra-cluster realignment process and normality statistical testing. We demonstrate improvements in documents with various levels of degradation, including recovered Uruguayan military archives and 17th to mid-20th century European newspapers.

Chaozhi Zhang, Wenxiang Ding, Roy Y. He et al.

The reconstruction of dynamic positron emission tomography (PET) images from noisy projection data is a significant but challenging problem. In this paper, we introduce an unsupervised learning approach, Non-negative Implicit Neural Representation Factorization (\texttt{NINRF}), based on low rank matrix factorization of unknown images and employing neural networks to represent both coefficients and bases. Mathematically, we demonstrate that if a sequence of dynamic PET images satisfies a generalized non-negative low-rank property, it can be decomposed into a set of non-negative continuous functions varying in the temporal-spatial domain. This bridges the well-established non-negative matrix factorization (NMF) with continuous functions and we propose using implicit neural representations (INRs) to connect matrix with continuous functions. The neural network parameters are obtained by minimizing the KL divergence, with additional sparsity regularization on coefficients and bases. Extensive experiments on dynamic PET reconstruction with Poisson noise demonstrate the effectiveness of the proposed method compared to other methods, while giving continuous representations for object's detailed geometric features and regional concentration variation.