Yuesheng Xu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Yuesheng Xu has a total of 130 co-authors. Here are the top co-authors by number of shared publications:
Charles A. Micchelli
17 shared publicationsZhongying Chen
15 shared publications- Lixin Shen
Department of Mathematics, Syracuse University, Syracuse, NY 13244, USA
11 shared publications - Rui Wang
School of Mathematics, Jilin University, Changchun, People’s Republic of China
10 shared publications - Charles Ross Schmidtlein
Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, USA
7 shared publications - Haizhang Zhang
School of Mathematics (Zhuhai), Sun Yat-Sen University, Zhuhai 519082, China
6 shared publications
Latest Publications
- Open Access
- 16 September 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 109 (2) , 50
The Adaptive Solution of High-Frequency Helmholtz Equations via Multi-Grade Deep Learning
The Helmholtz equation is fundamental for modeling wave propagation in acoustics, electromagnetics, and geophysics; however, high-frequency regimes remain notoriously difficult due to the severe numerical “pollution effect.” We propose FD-MGDL, an adaptive framework that synergizes finite difference discretizations with Multi-Grade Deep Learning (MGDL) to efficiently resolve high-frequency wavefields. Unlike standard physics-informed neural networks (PINNs), which frequently suffer from spectral bias and heavy automatic differentiation overhead, FD-MGDL employs a progressive, grade-wise training strategy that incrementally incorporates shallow sub-networks to refine residual errors. By leveraging ReLU activations in refinement grades, the framework reformulates the highly non-convex global optimization problem into a sequence of tractable convex subproblems, dramatically improving training stability and convergence reliability. Extensive numerical experiments in two and three dimensions with wavenumbers up to $$\kappa =200$$ demonstrate that FD-MGDL significantly outperforms single-grade networks and standard benchmark neural solvers in both accuracy and computational efficiency. When applied to an inhomogeneous concave velocity model, the proposed method accurately captures wave focusing and caustic formations, markedly outperforming classical five-point finite difference schemes in resolving sharp phase transitions and peak amplitudes. These results establish FD-MGDL as a robust, scalable, and mathematically grounded paradigm for high-frequency wave simulation in complex media.- 16 September 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 109 (2) , 50
Open AccessThe Adaptive Solution of High-Frequency Helmholtz Equations via Multi-Grade Deep Learning
The Helmholtz equation is fundamental for modeling wave propagation in acoustics, electromagnetics, and geophysics; however, high-frequency regimes remain notoriously difficult due to the severe numerical “pollution effect.” We propose FD-MGDL, an adaptive framework that synergizes finite difference discretizations with Multi-Grade Deep Learning (MGDL) to efficiently resolve high-frequency wavefields. Unlike standard physics-informed neural networks (PINNs), which frequently suffer from spectral bias and heavy automatic differentiation overhead, FD-MGDL employs a progressive, grade-wise training strategy that incrementally incorporates shallow sub-networks to refine residual errors. By leveraging ReLU activations in refinement grades, the framework reformulates the highly non-convex global optimization problem into a sequence of tractable convex subproblems, dramatically improving training stability and convergence reliability. Extensive numerical experiments in two and three dimensions with wavenumbers up to $$\kappa =200$$ demonstrate that FD-MGDL significantly outperforms single-grade networks and standard benchmark neural solvers in both accuracy and computational efficiency. When applied to an inhomogeneous concave velocity model, the proposed method accurately captures wave focusing and caustic formations, markedly outperforming classical five-point finite difference schemes in resolving sharp phase transitions and peak amplitudes. These results establish FD-MGDL as a robust, scalable, and mathematically grounded paradigm for high-frequency wave simulation in complex media. - Open Access
- 30 July 2026
- Published by Springer Nature in Journal of Mathematical Imaging and Vision
- Vol. 68 (4) , 48
MG-SpaIR: Multi-Grade Sparse-Guided Implicit Representation for Training-Data-Free Image Restoration
MG-SpaIR is a training-data-free framework for restoring a clean image from a single observation corrupted by a mixture of blur, downsampling, noise, and missing pixels. Building on implicit neural representations (INRs), we introduce a multi-grade residual hierarchy that progressively refines the reconstruction from low to high spatial frequencies across grades, improving representational fidelity and mitigating spectral limitations. To stabilize reconstruction optimization and suppress INR-induced artifacts, we further propose an explicit sparse proximal regularization (e.g., $$\ell _0$$ type) applied directly in the high-resolution image domain, which discourages spurious high-frequency patterns while preserving sharp structures. The resulting optimization is solved efficiently via a multi-grade proximal alternating scheme, and we establish convergence guarantees for the associated updates under standard regularity conditions. Experiments on mixed-degradation benchmarks demonstrate that MG-SpaIR consistently outperforms strong training-data-free baselines such as Deep Image Prior, providing a stable, interpretable, and data-efficient alternative to conventional learning-based restoration methods.- 30 July 2026
- Published by Springer Nature in Journal of Mathematical Imaging and Vision
- Vol. 68 (4) , 48
Open AccessMG-SpaIR: Multi-Grade Sparse-Guided Implicit Representation for Training-Data-Free Image Restoration
MG-SpaIR is a training-data-free framework for restoring a clean image from a single observation corrupted by a mixture of blur, downsampling, noise, and missing pixels. Building on implicit neural representations (INRs), we introduce a multi-grade residual hierarchy that progressively refines the reconstruction from low to high spatial frequencies across grades, improving representational fidelity and mitigating spectral limitations. To stabilize reconstruction optimization and suppress INR-induced artifacts, we further propose an explicit sparse proximal regularization (e.g., $$\ell _0$$ type) applied directly in the high-resolution image domain, which discourages spurious high-frequency patterns while preserving sharp structures. The resulting optimization is solved efficiently via a multi-grade proximal alternating scheme, and we establish convergence guarantees for the associated updates under standard regularity conditions. Experiments on mixed-degradation benchmarks demonstrate that MG-SpaIR consistently outperforms strong training-data-free baselines such as Deep Image Prior, providing a stable, interpretable, and data-efficient alternative to conventional learning-based restoration methods. - 4 March 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 107 (1) , 14
Sparsity-Guided Multi-Parameter Selection in $\ell _1$-Regularized Models via a Fixed-Point Proximity Approach- 4 March 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 107 (1) , 14
Sparsity-Guided Multi-Parameter Selection in $\ell _1$-Regularized Models via a Fixed-Point Proximity Approach- 4 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 45 (6) , 3246-3261
Computed Quantitative Planar Imaging for Targeted Alpha Therapy: Model-Based Sparse Reconstruction Validated With a Novel 225 Ac Epoxy Phantom
Targeted Alpha Therapy (TAT), using alpha-emitting radionuclides (AER) such as 225Ac, shows promise for the treatment of advanced and refractory cancers. Currently, TAT is prescribed on the basis of activity (e.g., MBq, kBq/kg), with no account taken of individual biodistribution or kinetics. The delivery of patient-specific treatment, based on absorbed dose criteria, requires in-vivo imaging of the AER biodistribution, a challenging scenario due to the scarcity of imageable photons. To address this, we present a novel computed quantitative planar (CQP) imaging method that reconstructs a coronal projection of the 3D AER distribution from anterior/posterior scintigraphy coregistered with CT. The model is regularized using maximum a posteriori estimation with sparse ℓ1 tight-framelet transforms and solved via a convergence-guaranteed fixed-point proximity algorithm. To experimentally evaluate our approach, we built a modular slab phantom containing a known distribution of 225Ac vitrified in epoxy. CQP reconstruction was characterized by significantly reduced bias and noise, improved spatial resolution, and better signal-to-noise ratios, compared to geometric mean methods. The CQP approach is clinically implementable with conventional SPECT/CT systems, without need for hardware additions or modifications, and can assist dosimetry workflows, especially where 3D SPECT/PET is impractical.
- 4 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Imaging
- Vol. 45 (6) , 3246-3261
Computed Quantitative Planar Imaging for Targeted Alpha Therapy: Model-Based Sparse Reconstruction Validated With a Novel 225 Ac Epoxy Phantom
Targeted Alpha Therapy (TAT), using alpha-emitting radionuclides (AER) such as 225Ac, shows promise for the treatment of advanced and refractory cancers. Currently, TAT is prescribed on the basis of activity (e.g., MBq, kBq/kg), with no account taken of individual biodistribution or kinetics. The delivery of patient-specific treatment, based on absorbed dose criteria, requires in-vivo imaging of the AER biodistribution, a challenging scenario due to the scarcity of imageable photons. To address this, we present a novel computed quantitative planar (CQP) imaging method that reconstructs a coronal projection of the 3D AER distribution from anterior/posterior scintigraphy coregistered with CT. The model is regularized using maximum a posteriori estimation with sparse ℓ1 tight-framelet transforms and solved via a convergence-guaranteed fixed-point proximity algorithm. To experimentally evaluate our approach, we built a modular slab phantom containing a known distribution of 225Ac vitrified in epoxy. CQP reconstruction was characterized by significantly reduced bias and noise, improved spatial resolution, and better signal-to-noise ratios, compared to geometric mean methods. The CQP approach is clinically implementable with conventional SPECT/CT systems, without need for hardware additions or modifications, and can assist dosimetry workflows, especially where 3D SPECT/PET is impractical.
- Open Access
The Adaptive Solution of High-Frequency Helmholtz Equations via Multi-Grade Deep Learning
The Helmholtz equation is fundamental to wave modeling in acoustics, electromagnetics, and seismic imaging, yet high-frequency regimes remain challenging due to the ``pollution effect''. We propose FD-MGDL, an adaptive framework integrating finite difference schemes with Multi-Grade Deep Learning to efficiently resolve high-frequency solutions. While traditional PINNs struggle with spectral bias and automatic differentiation overhead, FD-MGDL employs a progressive training strategy, incrementally adding hidden layers to refine the solution and maintain stability. Crucially, when using ReLU activation, our algorithm recasts the highly non-convex training problem into a sequence of convex subproblems. Numerical experiments in 2D and 3D with wavenumbers up to $κ=200$ show that FD-MGDL significantly outperforms single-grade and conventional neural solvers in accuracy and speed. Applied to an inhomogeneous concave velocity model, the framework accurately resolves wave focusing and caustics, surpassing the 5-point finite difference method in capturing sharp phase transitions and amplitude spikes. These results establish FD-MGDL as a robust, scalable solver for high-frequency wave equations in complex domains.Open AccessThe Adaptive Solution of High-Frequency Helmholtz Equations via Multi-Grade Deep Learning
The Helmholtz equation is fundamental to wave modeling in acoustics, electromagnetics, and seismic imaging, yet high-frequency regimes remain challenging due to the ``pollution effect''. We propose FD-MGDL, an adaptive framework integrating finite difference schemes with Multi-Grade Deep Learning to efficiently resolve high-frequency solutions. While traditional PINNs struggle with spectral bias and automatic differentiation overhead, FD-MGDL employs a progressive training strategy, incrementally adding hidden layers to refine the solution and maintain stability. Crucially, when using ReLU activation, our algorithm recasts the highly non-convex training problem into a sequence of convex subproblems. Numerical experiments in 2D and 3D with wavenumbers up to $κ=200$ show that FD-MGDL significantly outperforms single-grade and conventional neural solvers in accuracy and speed. Applied to an inhomogeneous concave velocity model, the framework accurately resolves wave focusing and caustics, surpassing the 5-point finite difference method in capturing sharp phase transitions and amplitude spikes. These results establish FD-MGDL as a robust, scalable solver for high-frequency wave equations in complex domains. - Open Access
- 31 January 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 106 (3) , 64
Adaptive Multi-Grade Deep Learning for Highly Oscillatory Fredholm Integral Equations of the Second Kind
This paper studies the use of Multi-Grade Deep Learning (MGDL) for solving highly oscillatory Fredholm integral equations of the second kind. We provide rigorous error analyses of continuous and discrete MGDL models, showing that the discrete model retains the convergence and stability of its continuous counterpart under sufficiently small quadrature error. We identify the DNN training error as the primary source of approximation error, motivating a novel adaptive MGDL algorithm that selects the network grade based on training performance. Numerical experiments with highly oscillatory (including wavenumber 500) and singular solutions confirm the accuracy, effectiveness and robustness of the proposed approach.
- 31 January 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 106 (3) , 64
Open AccessAdaptive Multi-Grade Deep Learning for Highly Oscillatory Fredholm Integral Equations of the Second Kind
This paper studies the use of Multi-Grade Deep Learning (MGDL) for solving highly oscillatory Fredholm integral equations of the second kind. We provide rigorous error analyses of continuous and discrete MGDL models, showing that the discrete model retains the convergence and stability of its continuous counterpart under sufficiently small quadrature error. We identify the DNN training error as the primary source of approximation error, motivating a novel adaptive MGDL algorithm that selects the network grade based on training performance. Numerical experiments with highly oscillatory (including wavenumber 500) and singular solutions confirm the accuracy, effectiveness and robustness of the proposed approach.
- Open Access
Adaptive Multi-Grade Deep Learning for Highly Oscillatory Fredholm Integral Equations of the Second Kind
This paper studies the use of Multi-Grade Deep Learning (MGDL) for solving highly oscillatory Fredholm integral equations of the second kind. We provide rigorous error analyses of continuous and discrete MGDL models, showing that the discrete model retains the convergence and stability of its continuous counterpart under sufficiently small quadrature error. We identify the DNN training error as the primary source of approximation error, motivating a novel adaptive MGDL algorithm that selects the network grade based on training performance. Numerical experiments with highly oscillatory (including wavenumber 500) and singular solutions confirm the accuracy, effectiveness and robustness of the proposed approach.
Open AccessAdaptive Multi-Grade Deep Learning for Highly Oscillatory Fredholm Integral Equations of the Second Kind
This paper studies the use of Multi-Grade Deep Learning (MGDL) for solving highly oscillatory Fredholm integral equations of the second kind. We provide rigorous error analyses of continuous and discrete MGDL models, showing that the discrete model retains the convergence and stability of its continuous counterpart under sufficiently small quadrature error. We identify the DNN training error as the primary source of approximation error, motivating a novel adaptive MGDL algorithm that selects the network grade based on training performance. Numerical experiments with highly oscillatory (including wavenumber 500) and singular solutions confirm the accuracy, effectiveness and robustness of the proposed approach.
- 21 August 2025
- Published by Elsevier in Neural Networks
- Vol. 193, 107995
Hypothesis spaces for deep learning
- 21 August 2025
- Published by Elsevier in Neural Networks
- Vol. 193, 107995
Hypothesis spaces for deep learning
- Open Access
Computational Advantages of Multi-Grade Deep Learning: Convergence Analysis and Performance Insights
Multi-grade deep learning (MGDL) has been shown to significantly outperform the standard single-grade deep learning (SGDL) across various applications. This work aims to investigate the computational advantages of MGDL focusing on its performance in image regression, denoising, and deblurring tasks, and comparing it to SGDL. We establish convergence results for the gradient descent (GD) method applied to these models and provide mathematical insights into MGDL's improved performance. In particular, we demonstrate that MGDL is more robust to the choice of learning rate under GD than SGDL. Furthermore, we analyze the eigenvalue distributions of the Jacobian matrices associated with the iterative schemes arising from the GD iterations, offering an explanation for MGDL's enhanced training stability.
Open AccessComputational Advantages of Multi-Grade Deep Learning: Convergence Analysis and Performance Insights
Multi-grade deep learning (MGDL) has been shown to significantly outperform the standard single-grade deep learning (SGDL) across various applications. This work aims to investigate the computational advantages of MGDL focusing on its performance in image regression, denoising, and deblurring tasks, and comparing it to SGDL. We establish convergence results for the gradient descent (GD) method applied to these models and provide mathematical insights into MGDL's improved performance. In particular, we demonstrate that MGDL is more robust to the choice of learning rate under GD than SGDL. Furthermore, we analyze the eigenvalue distributions of the Jacobian matrices associated with the iterative schemes arising from the GD iterations, offering an explanation for MGDL's enhanced training stability.
- 25 March 2025
- Published by World Scientific Pub Co Pte Ltd in Analysis and Applications
- Vol. 23 (08) , 1425-1467
Successive affine learning for deep neural networks
This paper introduces the Successive Affine Learning (SAL) model for constructing deep neural networks (DNNs). Traditionally, a DNN is trained by solving a single non-convex optimization problem, which is often challenging due to its high non-convexity and large number of layers. To address this challenge, the Multi-Grade Deep Learning (MGDL) model was recently initiated by the author of this paper, inspired by the human education system. MGDL constructs a DNN progressively, training a sequence of shallow networks with fewer layers. However, it still requires solving multiple non-convex optimization problems. The proposed SAL model evolves from MGDL by leveraging the structure of DNNs, where each layer consists of an affine transformation followed by an activation function. In SAL, we train the affine transformation first by solving a quadratic/convex optimization problem, incorporating the activation function only after learning the weight matrix and bias vector for the current layer. In the context of function approximation, given a target function, the SAL model generates an expansion of the function with adaptive basis functions in the form of DNNs. We establish key theoretical results, including the Pythagorean identity and the Parseval identity for the system generated by SAL. Furthermore, we prove a convergence theorem, showing that the SAL process either terminates after a finite number of grades or ensures a strictly decreasing sequence of optimal error norms, converging to a limit as the number of grades approaches infinity. Finally, we present numerical experiments as a proof of concept, demonstrating that the SAL model significantly outperforms traditional deep learning approaches
- 25 March 2025
- Published by World Scientific Pub Co Pte Ltd in Analysis and Applications
- Vol. 23 (08) , 1425-1467
Successive affine learning for deep neural networks
This paper introduces the Successive Affine Learning (SAL) model for constructing deep neural networks (DNNs). Traditionally, a DNN is trained by solving a single non-convex optimization problem, which is often challenging due to its high non-convexity and large number of layers. To address this challenge, the Multi-Grade Deep Learning (MGDL) model was recently initiated by the author of this paper, inspired by the human education system. MGDL constructs a DNN progressively, training a sequence of shallow networks with fewer layers. However, it still requires solving multiple non-convex optimization problems. The proposed SAL model evolves from MGDL by leveraging the structure of DNNs, where each layer consists of an affine transformation followed by an activation function. In SAL, we train the affine transformation first by solving a quadratic/convex optimization problem, incorporating the activation function only after learning the weight matrix and bias vector for the current layer. In the context of function approximation, given a target function, the SAL model generates an expansion of the function with adaptive basis functions in the form of DNNs. We establish key theoretical results, including the Pythagorean identity and the Parseval identity for the system generated by SAL. Furthermore, we prove a convergence theorem, showing that the SAL process either terminates after a finite number of grades or ensures a strictly decreasing sequence of optimal error norms, converging to a limit as the number of grades approaches infinity. Finally, we present numerical experiments as a proof of concept, demonstrating that the SAL model significantly outperforms traditional deep learning approaches