Research map / Architectures and efficiency / RNNs and normalization
Randomized Sparse Neural Galerkin Schemes for Solving Evolution Equations with Deep Networks
Jules Berman, Benjamin Peherstorfer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace PINN with RSNG (s=800) | RSNG · Allen-Cahn | L2 relative error: 6.85e-02 -> 1.66e-05 | Improved |
| Replace Causal PINN with RSNG (s=800) | RSNG · Allen-Cahn | L2 relative error: 3.84e-04 -> 1.66e-05 | Improved |
| Replace Causal PINN with RSNG (s=800) | RSNG · Burgers' | L2 relative error: 5.19e-04 -> 6.07e-05 | Improved |
| Replace Causal PINN with RSNG (s=125) | RSNG · Burgers' | training time (s): 23027 -> 188 | Improved |
| Reduce sparsity s from 800 to 150 in RSNG (ablation) | RSNG · Allen-Cahn | training time (s): 776 -> 63 | Improved |
| Reduce sparsity s from 800 to 150 in RSNG (ablation) | RSNG · Allen-Cahn | L2 relative error: 1.66e-05 -> 5.34e-05 | Hurt |
| Reduce sparsity s from 800 to 125 in RSNG (ablation) | RSNG · Burgers' | L2 relative error: 6.07e-05 -> 2.05e-04 | Hurt |
| Use sparse randomized updates (s below p) instead of dense updates in 7-layer network (ablation) | Neural Galerkin · Allen-Cahn, Burgers', Vlasov | relative error: 2--3x lower error | Improved |
| Use 7-layer network with sparse updates (s equal to 3-layer p) instead of 3-layer dense updates at equal cost (ablation) | Neural Galerkin · Allen-Cahn, Burgers', Vlasov | relative error | Improved |
| Use RSNG instead of dense direct solver | Neural Galerkin · Burgers' | runtime at comparable error | Improved |
| Use RSNG instead of dense iterative solver | Neural Galerkin · Burgers' | runtime at comparable error | Improved |
| Periodically refit network from random initialization (dense iterative solver) (ablation) | Neural Galerkin · Burgers' | relative error vs runtime | Mixed |
Closest papers by meaning
- PRDP: Progressively Refined Differentiable Physics (ICLR 2025 · similarity 0.86)
- Adversarial Adaptive Sampling: Unify PINN and Optimal Transport for the Approximation of PDEs (ICLR 2024 · similarity 0.86)
- RandNet-Parareal: a time-parallel PDE solver using Random Neural Networks (NeurIPS 2024 · similarity 0.86)
- Breakeven complexity: A new perspective on neural partial differential equation solvers (NeurIPS 2026 · similarity 0.86)
- Random Neural Network Expressivity for Non-Linear Partial Differential Equations (NeurIPS 2026 · similarity 0.86)
- S$^{2}$-PINN: Stochastic Separable Physics-Informed Neural Networks (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
