Yangshuai Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yangshuai Wang has a total of 121 co-authors. Here are the top co-authors by number of shared publications:
- Christoph Ortner
Department of Mathematics, University of British Columbia, 1984 Mathematics Road, Vancouver, BC V6T 1Z2, Canada
9 shared publications - Lei Zhang
School of Mathematical Sciences Shanghai Jiao Tong University 200240 Shanghai P. R. China
7 shared publications Cheuk Hin Ho
5 shared publicationsChristoph Ortner
Department of Mathematics, University of British Columbia, Vancouver, Canada
5 shared publications- Hao Wang
School of Mathematics Sichuan University No. 24 Yihuan Road Chengdu China
3 shared publications - James R. Kermode
University of Warwick
3 shared publications
Latest Publications
- 13 October 2026
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Scientific Computing
- Vol. 48 (5) , A2934-A2958
Higher-Order Boundary Conditions for Atomistic Dislocation Simulations
We present a higher-order boundary condition for atomistic simulations of dislocations that address the slow convergence of standard supercell methods. The method is based on a multipole expansion of the equilibrium displacement, combining continuum predictor solutions with discrete moment corrections. The continuum predictors are computed by solving a hierarchy of singular elliptic PDEs via a Galerkin spectral method, while moment coefficients are evaluated from force-moment identities with controlled approximation error. A key feature is the coupling between accurate continuum predictors and moment evaluations, enabling the construction of systematically improvable high-order boundary conditions. We thus design novel algorithms, and numerical results for screw and edge dislocations confirm the predicted convergence rates in geometry and energy norms, with reduced finite-size effects and moderate computational cost.
- 13 October 2026
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Scientific Computing
- Vol. 48 (5) , A2934-A2958
Higher-Order Boundary Conditions for Atomistic Dislocation Simulations
We present a higher-order boundary condition for atomistic simulations of dislocations that address the slow convergence of standard supercell methods. The method is based on a multipole expansion of the equilibrium displacement, combining continuum predictor solutions with discrete moment corrections. The continuum predictors are computed by solving a hierarchy of singular elliptic PDEs via a Galerkin spectral method, while moment coefficients are evaluated from force-moment identities with controlled approximation error. A key feature is the coupling between accurate continuum predictors and moment evaluations, enabling the construction of systematically improvable high-order boundary conditions. We thus design novel algorithms, and numerical results for screw and edge dislocations confirm the predicted convergence rates in geometry and energy norms, with reduced finite-size effects and moderate computational cost.
- Open Access
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Posterior-state mechanics digital twins: sensor-calibrated interface covariance transport across changing representations
Sensor agreement does not guarantee trustworthy full-field uncertainty when a mechanics digital twin changes its reduced representation. Scalar covariance inflation cannot distinguish interface discrepancy from error in field directions omitted by the new representation. We formulate a posterior-state twin that transports Gaussian field posteriors across blockwise reduced representations. Commissioning sensor innovations select a diagonal-plus-low-rank interface covariance by prequential likelihood. The covariance is applied in physical coordinates before projection and transferred by an information-equivalent inverse-square-root action. Thermal and three-dimensional finite-element studies test covariance misspecification, coefficient rotations and changing modal subspaces. On the cantilever, structured transport attains near-nominal coverage with 48% narrower intervals and 42% lower field error than a sensor-calibrated scalar model. A component-aware kernel also recovers displacement-component scaling from sensor data alone. When discrepancy occupies modes absent from the twin, structured transport exposes unresolved directions that scalar inflation masks through over-dispersion. Sensor-selected enrichment then reduces field error by 54% and interval width by 43% relative to a recalibrated fixed 30-mode modal Kalman filter while preserving mean coverage. A common-subspace control retains the smaller representation when discrepancy remains inside the shared modes. The low-rank implementation preserves dense Gaussian information at interfaces with thousands of degrees of freedom and makes blockwise transport practical. The recursion is exact for linear--Gaussian models. The resulting mechanics twin carries calibrated, scalable full-field uncertainty across representation changes.
- 22 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPosterior-state mechanics digital twins: sensor-calibrated interface covariance transport across changing representations
Sensor agreement does not guarantee trustworthy full-field uncertainty when a mechanics digital twin changes its reduced representation. Scalar covariance inflation cannot distinguish interface discrepancy from error in field directions omitted by the new representation. We formulate a posterior-state twin that transports Gaussian field posteriors across blockwise reduced representations. Commissioning sensor innovations select a diagonal-plus-low-rank interface covariance by prequential likelihood. The covariance is applied in physical coordinates before projection and transferred by an information-equivalent inverse-square-root action. Thermal and three-dimensional finite-element studies test covariance misspecification, coefficient rotations and changing modal subspaces. On the cantilever, structured transport attains near-nominal coverage with 48% narrower intervals and 42% lower field error than a sensor-calibrated scalar model. A component-aware kernel also recovers displacement-component scaling from sensor data alone. When discrepancy occupies modes absent from the twin, structured transport exposes unresolved directions that scalar inflation masks through over-dispersion. Sensor-selected enrichment then reduces field error by 54% and interval width by 43% relative to a recalibrated fixed 30-mode modal Kalman filter while preserving mean coverage. A common-subspace control retains the smaller representation when discrepancy remains inside the shared modes. The low-rank implementation preserves dense Gaussian information at interfaces with thousands of degrees of freedom and makes blockwise transport practical. The recursion is exact for linear--Gaussian models. The resulting mechanics twin carries calibrated, scalable full-field uncertainty across representation changes.
- 25 June 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 108 (2) , 57
Amplitude Expansion Phase Field Crystal (APFC) Modeling based Efficient Dislocation Simulations using Fourier Pseudospectral Method
- 25 June 2026
- Published by Springer Nature in Journal of Scientific Computing
- Vol. 108 (2) , 57
Amplitude Expansion Phase Field Crystal (APFC) Modeling based Efficient Dislocation Simulations using Fourier Pseudospectral Method
- Open Access
- 23 June 2026
- Published by Elsevier in SSRN Electronic Journal
Higher-Order Uncertainty Quantification for Physics-Informed Neural Networks
Physics-informed neural network (PINN) solvers increasingly attach uncertainty estimates to predicted fields and inferred parameters, and these estimates often determine whether a solution is accepted, an inverse result is trusted, or additional data are acquired. Most existing PINN uncertainty quantification addresses uncertainty in the predicted quantity; the stability of the uncertainty estimate itself under training variability is usually left unmeasured. We formulate Higher-Order Uncertainty Quantification (HOUQ) for PINNs by shifting the target of UQ from the predicted quantity to the reported uncertainty estimate. Given a baseline UQ mechanism, HOUQ recomputes its uncertainty estimate under admissible perturbations, including initialization, collocation resampling, and loss-weight changes, and converts the resulting variability into field-wise reliability maps, parameter-interval drift, and decision-level risk scores. Numerical evidence from forward Burgers, Poisson, and Helmholtz benchmarks and inverse diffusion, reaction-diffusion, Poisson, and Allen--Cahn problems shows that HOUQ separates uncertainty magnitude from uncertainty reliability. In a two-parameter inverse reaction-diffusion study, validation-selected UQ/HOUQ scores rank high-risk runs before retraining and guide targeted observations that reduce field RMSE by 43.0% and parameter error by 13.5%. These results establish HOUQ as a second-level reliability measure for PINN uncertainty analysis and connect reported uncertainty to computational PDE decisions.
- 23 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessHigher-Order Uncertainty Quantification for Physics-Informed Neural Networks
Physics-informed neural network (PINN) solvers increasingly attach uncertainty estimates to predicted fields and inferred parameters, and these estimates often determine whether a solution is accepted, an inverse result is trusted, or additional data are acquired. Most existing PINN uncertainty quantification addresses uncertainty in the predicted quantity; the stability of the uncertainty estimate itself under training variability is usually left unmeasured. We formulate Higher-Order Uncertainty Quantification (HOUQ) for PINNs by shifting the target of UQ from the predicted quantity to the reported uncertainty estimate. Given a baseline UQ mechanism, HOUQ recomputes its uncertainty estimate under admissible perturbations, including initialization, collocation resampling, and loss-weight changes, and converts the resulting variability into field-wise reliability maps, parameter-interval drift, and decision-level risk scores. Numerical evidence from forward Burgers, Poisson, and Helmholtz benchmarks and inverse diffusion, reaction-diffusion, Poisson, and Allen--Cahn problems shows that HOUQ separates uncertainty magnitude from uncertainty reliability. In a two-parameter inverse reaction-diffusion study, validation-selected UQ/HOUQ scores rank high-risk runs before retraining and guide targeted observations that reduce field RMSE by 43.0% and parameter error by 13.5%. These results establish HOUQ as a second-level reliability measure for PINN uncertainty analysis and connect reported uncertainty to computational PDE decisions.
- Open Access
- 17 June 2026
- Published by Elsevier in SSRN Electronic Journal
Liquid Random Feature Methods for Time-Dependent Partial Differential Equations
A central challenge in mesh-free space--time approximation for time-dependent partial differential equations is to represent evolving temporal scales while keeping residual minimization computationally tractable. Random feature methods simplify this algebraic problem by freezing nonlinear trial functions and fitting only a linear readout, but standard static space--time activations provide no explicit relaxation-scale mechanism, making temporal-scale resolution a finite-dimensional bottleneck in stiff, dispersive, or multi-scale regimes. We introduce liquid random feature methods (L-RFM), which replace static temporal activations by closed-form liquid time-constant responses with sampled relaxation scales. The resulting frozen features form temporally structured local or global trial spaces with analytic space--time derivatives for residual least-squares assembly. A density theorem proves density of the deterministic trial spaces in the continuous space--time function class, and a temporal-rank calculation clarifies the role of sampled relaxation scales. Ablation and finite-feature tests identify the liquid temporal response as the primary source of the observed accuracy improvement. Across stiff reaction--diffusion, nonlinear transport, dispersive, complex-valued, and multidimensional benchmarks, L-RFM improves finite-feature accuracy in regimes where temporal-scale representation controls the approximation. By embedding relaxation scales directly into frozen trial functions, L-RFM provides a route to high-accuracy continuous space--time surrogates for evolutionary PDEs while preserving the simplicity of linear least-squares solvers.
- 17 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessLiquid Random Feature Methods for Time-Dependent Partial Differential Equations
A central challenge in mesh-free space--time approximation for time-dependent partial differential equations is to represent evolving temporal scales while keeping residual minimization computationally tractable. Random feature methods simplify this algebraic problem by freezing nonlinear trial functions and fitting only a linear readout, but standard static space--time activations provide no explicit relaxation-scale mechanism, making temporal-scale resolution a finite-dimensional bottleneck in stiff, dispersive, or multi-scale regimes. We introduce liquid random feature methods (L-RFM), which replace static temporal activations by closed-form liquid time-constant responses with sampled relaxation scales. The resulting frozen features form temporally structured local or global trial spaces with analytic space--time derivatives for residual least-squares assembly. A density theorem proves density of the deterministic trial spaces in the continuous space--time function class, and a temporal-rank calculation clarifies the role of sampled relaxation scales. Ablation and finite-feature tests identify the liquid temporal response as the primary source of the observed accuracy improvement. Across stiff reaction--diffusion, nonlinear transport, dispersive, complex-valued, and multidimensional benchmarks, L-RFM improves finite-feature accuracy in regimes where temporal-scale representation controls the approximation. By embedding relaxation scales directly into frozen trial functions, L-RFM provides a route to high-accuracy continuous space--time surrogates for evolutionary PDEs while preserving the simplicity of linear least-squares solvers.
- 4 May 2026
- Published by Elsevier in Journal of Computational Physics
- Vol. 561
A conformal prediction framework for uncertainty quantification in physics-informed neural networks
- 4 May 2026
- Published by Elsevier in Journal of Computational Physics
- Vol. 561
A conformal prediction framework for uncertainty quantification in physics-informed neural networks
- Open Access
- 27 April 2026
- Published by Springer Nature in npj Computational Materials
- Vol. 12 (1) , 225
Flexible uncertainty calibration for machine-learned interatomic potentials
Reliable uncertainty quantification (UQ) is essential for developing machine-learned interatomic potentials (MLIPs) in predictive atomistic simulations. Conformal prediction (CP) is a statistical framework that constructs prediction intervals with guaranteed coverage under minimal assumptions, making it an attractive tool for UQ. However, existing CP techniques, while offering formal coverage guarantees, often lack accuracy, scalability, and adaptability to the complexity of atomic environments. In this work, we present a flexible uncertainty calibration framework for MLIPs, inspired by CP but reformulated as a parameterized optimization problem. This formulation enables the direct learning of environment-dependent quantile functions, producing sharper and more adaptive predictive intervals at negligible computational cost. Using the foundation model MACE-MP-0 as a representative case, we demonstrate the framework across diverse benchmarks, including ionic crystals, catalytic surfaces, and molecular systems. Our results achieve substantial improvements in uncertainty-error correlation, improve the detection of high-error configurations for active learning, and transfer reliably across distinct exchange-correlation functionals. Importantly, it is general, data efficient, and compatible with diverse MLIP architectures and baseline UQ schemes, offering a practical route toward robust and transferable atomistic simulations.
- 27 April 2026
- Published by Springer Nature in npj Computational Materials
- Vol. 12 (1) , 225
Open AccessFlexible uncertainty calibration for machine-learned interatomic potentials
Reliable uncertainty quantification (UQ) is essential for developing machine-learned interatomic potentials (MLIPs) in predictive atomistic simulations. Conformal prediction (CP) is a statistical framework that constructs prediction intervals with guaranteed coverage under minimal assumptions, making it an attractive tool for UQ. However, existing CP techniques, while offering formal coverage guarantees, often lack accuracy, scalability, and adaptability to the complexity of atomic environments. In this work, we present a flexible uncertainty calibration framework for MLIPs, inspired by CP but reformulated as a parameterized optimization problem. This formulation enables the direct learning of environment-dependent quantile functions, producing sharper and more adaptive predictive intervals at negligible computational cost. Using the foundation model MACE-MP-0 as a representative case, we demonstrate the framework across diverse benchmarks, including ionic crystals, catalytic surfaces, and molecular systems. Our results achieve substantial improvements in uncertainty-error correlation, improve the detection of high-error configurations for active learning, and transfer reliably across distinct exchange-correlation functionals. Importantly, it is general, data efficient, and compatible with diverse MLIP architectures and baseline UQ schemes, offering a practical route toward robust and transferable atomistic simulations.
- Open Access
- 1 March 2026
- Published by IOP Publishing in AI for Science
- Vol. 2 (1) , 015004
Beyond Adam: disentangling optimizer effects in the fine-tuning of atomistic foundation models
Atomistic foundation models constitute a paradigm shift in computational materials science by providing universal machine-learned interatomic potentials with broad transferability across chemical spaces. Although fine-tuning is essential for adapting these pretrained models to specific target systems, the influence of the optimization algorithm on this process remains insufficiently characterized. In this work, we perform a rigorous benchmark of seven first-order optimizers, including Adam, AdamW, rectified Adam, stochastic gradient descent (SGD), layerwise adaptive moment, Ranger, and ScheduleFree, for the fine-tuning of foundation models across molecular, crystalline, and liquid regimes. We evaluate these algorithms based on energy and force accuracy for both in-distribution and out-of-distribution configurations, as well as their impact on downstream physical properties such as elastic moduli, phonon spectra, and interfacial dynamics. We interpret these empirical results through a preconditioning framework that views each optimizer as a data-dependent linear transformation of the gradient. This analysis clarifies how different update rules impose specific spectral filters on the effective loss Hessian. Across all regimes, AdamW and ScheduleFree achieve superior curvature conditioning and force accuracy, whereas SGD exhibits slow convergence and instability. Furthermore, we demonstrate that a brief second-order refinement stage reduces residual anisotropy in the loss landscape and enhances the fidelity of physical observables without increasing inference costs. These findings provide conceptual insight and practical guidance for selecting and designing optimizers to ensure the stable and efficient fine-tuning of universal interatomic potentials.
- 1 March 2026
- Published by IOP Publishing in AI for Science
- Vol. 2 (1) , 015004
Open AccessBeyond Adam: disentangling optimizer effects in the fine-tuning of atomistic foundation models
Atomistic foundation models constitute a paradigm shift in computational materials science by providing universal machine-learned interatomic potentials with broad transferability across chemical spaces. Although fine-tuning is essential for adapting these pretrained models to specific target systems, the influence of the optimization algorithm on this process remains insufficiently characterized. In this work, we perform a rigorous benchmark of seven first-order optimizers, including Adam, AdamW, rectified Adam, stochastic gradient descent (SGD), layerwise adaptive moment, Ranger, and ScheduleFree, for the fine-tuning of foundation models across molecular, crystalline, and liquid regimes. We evaluate these algorithms based on energy and force accuracy for both in-distribution and out-of-distribution configurations, as well as their impact on downstream physical properties such as elastic moduli, phonon spectra, and interfacial dynamics. We interpret these empirical results through a preconditioning framework that views each optimizer as a data-dependent linear transformation of the gradient. This analysis clarifies how different update rules impose specific spectral filters on the effective loss Hessian. Across all regimes, AdamW and ScheduleFree achieve superior curvature conditioning and force accuracy, whereas SGD exhibits slow convergence and instability. Furthermore, we demonstrate that a brief second-order refinement stage reduces residual anisotropy in the loss landscape and enhances the fidelity of physical observables without increasing inference costs. These findings provide conceptual insight and practical guidance for selecting and designing optimizers to ensure the stable and efficient fine-tuning of universal interatomic potentials.
- 18 February 2026
- Published by Elsevier in Computer Physics Communications
- Vol. 322
Adaptive quasicontinuum methods and simulations for crystal defects with a theory based unified a posteriori error estimate
- 18 February 2026
- Published by Elsevier in Computer Physics Communications
- Vol. 322
Adaptive quasicontinuum methods and simulations for crystal defects with a theory based unified a posteriori error estimate
- Top Cited PapersOpen Access
- 26 January 2026
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 139 (4)
Fine-tuning universal machine-learned interatomic potentials: A Tutorial on methods and applications
Universal machine-learned interatomic potentials (U-MLIPs) have demonstrated broad applicability across diverse atomistic systems but often require fine-tuning to achieve task-specific accuracy. While the number of available U-MLIPs and their fine-tuning applications are rapidly expanding, there remains a lack of systematic guidance on how to effectively fine-tune these models. This Tutorial provides a comprehensive, step-by-step guide to fine-tuning U-MLIPs for computational materials modeling. Using the recently released MACE-MP-0 as a representative foundation model, we illustrate the full workflow of data set preparation, hyperparameter selection, model training, and validation. Beyond methodological guidance, we conduct systematic case studies on solid-state electrolytes, stacking fault defects in metals, semiconductors, solid–liquid interfacial interactions in low-dimensional systems, and more complicated heterointerfaces. These examples demonstrate that fine-tuning substantially improves predictive accuracy while maintaining affordable computational cost, accelerates training convergence, enhances out-of-distribution generalization, and achieves superior data efficiency. Remarkably, fine-tuned foundation models can even capture aspects of long-range physics without explicit corrections. Together, these results highlight that fine-tuning not only provides a practical recipe for applying U-MLIPs but also offers new insights into their physical fidelity and potential for advancing large-scale atomistic simulations. To support practical applications, we include code examples that enable researchers, particularly those new to the field, to efficiently incorporate fine-tuned U-MLIPs into their workflows.
- 26 January 2026
- Published by AIP Publishing in Journal of Applied Physics
- Vol. 139 (4)
Top Cited PapersOpen AccessFine-tuning universal machine-learned interatomic potentials: A Tutorial on methods and applications
Universal machine-learned interatomic potentials (U-MLIPs) have demonstrated broad applicability across diverse atomistic systems but often require fine-tuning to achieve task-specific accuracy. While the number of available U-MLIPs and their fine-tuning applications are rapidly expanding, there remains a lack of systematic guidance on how to effectively fine-tune these models. This Tutorial provides a comprehensive, step-by-step guide to fine-tuning U-MLIPs for computational materials modeling. Using the recently released MACE-MP-0 as a representative foundation model, we illustrate the full workflow of data set preparation, hyperparameter selection, model training, and validation. Beyond methodological guidance, we conduct systematic case studies on solid-state electrolytes, stacking fault defects in metals, semiconductors, solid–liquid interfacial interactions in low-dimensional systems, and more complicated heterointerfaces. These examples demonstrate that fine-tuning substantially improves predictive accuracy while maintaining affordable computational cost, accelerates training convergence, enhances out-of-distribution generalization, and achieves superior data efficiency. Remarkably, fine-tuned foundation models can even capture aspects of long-range physics without explicit corrections. Together, these results highlight that fine-tuning not only provides a practical recipe for applying U-MLIPs but also offers new insights into their physical fidelity and potential for advancing large-scale atomistic simulations. To support practical applications, we include code examples that enable researchers, particularly those new to the field, to efficiently incorporate fine-tuned U-MLIPs into their workflows.