The curvature-aware precision controller adapts between FP32 and FP64 during PINN training to match double-precision accuracy at reduced computational cost.
Self-adaptive loss balanced Physics -informed neural networks
4 Pith papers cite this work, alongside 435 external citations. Polarity classification is still indexing.
4
Pith papers citing it
435
external citations · Crossref
representative citing papers
NeuroSPICE uses PINNs to solve circuit DAEs via residual minimization, creating surrogate models for optimization and handling nonlinear emerging devices.
Switching between coarse and fine collocation datasets during PINN training, like a multigrid V-cycle, improves accuracy by up to 64 percent on lid-driven cavity benchmarks.
A DeepONet-PINN hybrid with latent perturbation is proposed for PDE parameter estimation, but the practical loss is not the derived variational objective.
citing papers explorer
-
Curvature-aware dynamic precision approach for physics-informed neural networks
The curvature-aware precision controller adapts between FP32 and FP64 during PINN training to match double-precision accuracy at reduced computational cost.
-
Physics-Informed Neural Networks for Device and Circuit Modeling: A Case Study of NeuroSPICE
NeuroSPICE uses PINNs to solve circuit DAEs via residual minimization, creating surrogate models for optimization and handling nonlinear emerging devices.
-
Multi-level datasets training method in Physics-Informed Neural Networks
Switching between coarse and fine collocation datasets during PINN training, like a multigrid V-cycle, improves accuracy by up to 64 percent on lid-driven cavity benchmarks.
-
Deep Operator Networks for Bayesian Parameter Estimation in PDEs
A DeepONet-PINN hybrid with latent perturbation is proposed for PDE parameter estimation, but the practical loss is not the derived variational objective.