Low-L2-error neural operators for PDE warm-starts can yield indefinite Jacobians; energy-penalized fine-tuning restores positive definiteness and achieves up to 5.4x speedup on 6.4M DOF hyperelasticity problems.
Résolution d' EDP par un schéma en temps «pararéel »
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
representative citing papers
The parareal algorithm is shown to converge linearly for semilinear parabolic PDEs with H^2 initial data by using stable rational approximations and first-order linearization as coarse propagators, with a sharp convergence factor estimate.
A parallel two-level stepping method for athermal quasistatic deformation achieves average computational speed-ups of 2.02 to 6.33 times with 4 to 32 threads.
The serial scaling hypothesis formalizes inherently serial problems in complexity theory and demonstrates that diffusion models cannot solve them.
citing papers explorer
-
Spectrally Safe Neural Operator Warm-Starts for Large-Scale Newton Solvers
Low-L2-error neural operators for PDE warm-starts can yield indefinite Jacobians; energy-penalized fine-tuning restores positive definiteness and achieves up to 5.4x speedup on 6.4M DOF hyperelasticity problems.
-
Linear Convergence of Parareal Algorithm for Semilinear Parabolic Equations
The parareal algorithm is shown to converge linearly for semilinear parabolic PDEs with H^2 initial data by using stable rational approximations and first-order linearization as coarse propagators, with a sharp convergence factor estimate.