This paper develops isotonic Bayesian surrogate modeling for computer experiments involving many input variables and complex physical phenomena, including particle collisions and rocket propulsion. The method incorporates monotonicity or isotonicity information elicited from physical systems to improve surrogate modeling when the available sample size is limited, while retaining uncertainty quantification.
See LICENSE fileBayesian surrogate modelingIsotonic modelingMonotonicity constraints
KernelOnet is an interpretable neural operator framework that incorporates kernel functions explicitly into its architecture, aligning the operator structure with boundary-type kernel-expansion forms for physics modeling.
See LICENSE fileneural operatorskernel functionsinterpretable machine learning
Latent Twin Operator is a latent-space surrogate for time-evolving PDEs that uses a ConvCNP-style encoder and decoder to process sensor observations at arbitrary placements and support queries at any resolution.
See LICENSE filelatent-space surrogatetime-evolving PDEsconditional neural process
This paper presents an LSTM-based pressure predictive control method for motion control of a vertebraic soft robotic tail. The approach targets nonlinear material behavior and complex continuum kinematics, and coordinates the tail with a quadruped robot.
See LICENSE filesoft roboticspressure predictive controlLSTM
This paper studies the structural, thermophysical, and dynamical properties of MgCl2 molten salt over a wide temperature range using AIMD and molecular dynamics simulations driven by machine-learning interaction potentials. The MLIP approach is reported to achieve near-AIMD accuracy at a fraction of the computational cost, enabling larger systems and longer timescales.
See LICENSE fileMgCl2 molten saltMolecular dynamicsMachine-learning interaction potentials
NHMO is a neural solver for elliptic PDEs on variable-shape domains that learns harmonic measure using a transformer-based boundary kernel supervised by Walk-on-Spheres exit samples.
See LICENSE fileElliptic PDEHarmonic MeasureTransformer Boundary Kernel
This paper studies predictive representation pretraining for parametric partial differential equations, modeling latent dynamics as an alternative to reconstruction-based learning of physical fields.
See LICENSE fileParametric PDEsLatent DynamicsPredictive Representation Learning
This study develops a PINN framework for the Savage-Hutter model of depth-averaged granular flow to support mountain avalanche hazard assessment. It advances from one-dimensional analytical verification of three problems to two-dimensional experimental validation of avalanche dynamics.
See LICENSE filephysics-informed neural networksavalanche dynamicsdepth-averaged granular flow
This work develops a physics-informed, self-supervised framework for generating digital brain MRI phantoms from conventional weighted MR images without ground-truth parametric maps or anatomical segmentation. It predicts T1, T2, and proton-density maps and reconstructs the input images through a differentiable MRI signal model.
See LICENSE fileDigital Brain MRI PhantomsPhysics-Informed Machine LearningSelf-Supervised Learning
POIL uses a shared representation to transfer demonstrated manipulation trajectories to novel objects and execute them robustly across changing scenes, grasp configurations, and external disturbances.
See LICENSE fileRobotic ManipulationOne-Shot Imitation LearningStable Dynamical Systems
This paper studies preconditioned training for physics-informed neural operators across general physics simulations, using governing equations rather than solver-generated datasets to reduce offline data-generation cost and enable fresh samples during optimization.
See LICENSE filephysics-informed neural operatorpreconditioned trainingscientific machine learning
This study evaluates multi-horizon forecasting for the Lomnicky Stit neutron monitor time series in geophysics, comparing seasonal baselines, modern deep sequence models, and functional quantum-inspired architectures.
See LICENSE fileneutron monitoringtime series forecastinggeophysics
SCRINN addresses generic partial differential equations and function approximation through structured sparse connectivity and an orthogonality- and normalization-guided loss that promotes an approximately orthonormal neural basis and progressive evaluation-matrix rank growth.
See LICENSE fileScientific Machine LearningPartial Differential EquationsSparse Neural Networks
This paper introduces a representation curriculum for physics-informed neural networks, implemented through staged depth training that learns representations independently of their predictors and progressively refines them.
See LICENSE filephysics-informed neural networksrepresentation curriculumstaged depth training
This paper develops a tensor-train compressed separable physics-informed neural network framework for high-dimensional parametric PDEs. It uses a Gauss-Newton pullback metric to provide an operator-informed curvature notion and connects second-order PINN optimization with natural-gradient schemes.
See LICENSE filephysics-informed neural networkshigh-dimensional parametric PDEssecond-order optimization
The paper introduces U(1) gauge-equivariant representation learning for mean-field Bloch states, explicitly addressing the gauge redundancy of electronic states in first-principles condensed-matter systems.
See LICENSE fileBloch state representationGauge-equivariant learningCondensed-matter electronic structure
This paper studies short-term forecasting of cloud-induced power ramps across distributed photovoltaic and irradiance sensors using an advection-aware graph neural network and a self-supervised cloud-motion estimator.
See LICENSE filesolar power forecastinggraph neural networkcloud motion estimation
CrystAF studies where physics should enter all-atom molecular crystal generation by comparing training, post-training, and inference strategies with the generator and physical signal held fixed.
See LICENSE fileMolecular crystalsGenerative modelingInteratomic potential
This paper evaluates physics-informed residual constraints in neural PDE surrogates for nonlinear dynamics, finding that known-equation residuals outperform the strongest matched generic regularizer under equal tuning budgets.
See LICENSE fileneural PDE surrogatesresidual constraintsnonlinear dynamics
This paper presents a two-stage reduction pipeline for parametrised PDE shape problems, combining Active Subspace-guided structured control-point selection with Sinkhorn autoencoders for free-form deformation and reduced-order modelling.
See LICENSE fileParametrised PDEsShape OptimisationFree-Form Deformation