Unsupervised Recurrent Graph Neural Networks Accelerate Plasma Simulation Poisson Solvers
Researchers have developed an unsupervised recurrent Graph Neural Network that trains online to generate accurate initial guesses for iterative linear ...
Researchers have developed an unsupervised recurrent Graph Neural Network that trains online to generate accurate initial guesses for iterative linear ...
A new Bayesian analysis of charged-particle data across four different nuclear collision systems establishes a universal energy-loss scaling exponent consistent ...
A research team has developed a new finite-volume scheme to calculate the step-scaling function in Hamiltonian lattice gauge theory, enabling ...
Researchers have introduced an adaptive weight-loss gradient-augmented physics-informed neural network that significantly improves the accuracy and stability of computing soliton ...
A novel stochastic path sampler enhances lattice field theory simulations by using nonequilibrium thermodynamics to accelerate computation.
Researchers have developed modified finite renormalization schemes for quasi-parton distribution functions to eliminate dependence on reference momentum.
A new physics-aware neural network framework automates the calibration of carbon nanotube field-effect transistor models by integrating silicon TCAD references ...
Researchers have developed WorldParticle, a unified deep-learning architecture that simulates diverse physical phenomena by predicting Lagrangian particle movements step by ...
Researchers have demonstrated a scalable time-series framework using Haar wavelets and exact FAISS indexing that retrieves physical structural patterns across ...
Researchers developed a weak-form discrete physical-residual physics-informed neural network framework that eliminates high-order spatial automatic differentiation to solve complex multigroup ...
© 2026 Scienmag - Science Magazine
© 2026 Scienmag - Science Magazine