Research map / Optimization / Optimizers
Achieving High Accuracy with PINNs via Energy Natural Gradient Descent
Johannes Müller, Marius Zeinhofer
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 |
|---|---|---|---|
| Use energy natural gradient descent instead of GD | PINN for 2D Poisson equation · 2D Poisson equation | relative L2 error (median): 8.2 \cdot 10^{-3} -> 2.4\cdot 10^{-7} | Improved |
| Use energy natural gradient descent instead of Adam | PINN for 1D heat equation · 1D heat equation | relative L2 error (median): 1.0\cdot10^{-3} -> 6.3\cdot 10^{-6} | Improved |
| Use energy natural gradient descent instead of BFGS | PINN for 2D Poisson equation · 2D Poisson equation | relative L2 error (median): 4.4 \cdot 10^{-4} -> 2.4\cdot 10^{-7} | Improved |
| Use Hilbert natural gradient instead of energy natural gradient | PINN for 2D Poisson equation · 2D Poisson equation | relative L2 error (median): 2.4\cdot 10^{-7} -> 1.2 | Hurt |
| Use energy natural gradient descent instead of Adam | Deep Ritz method for nonlinear elliptic equation · 1D nonlinear elliptic equation (deep Ritz) | relative L2 error (median): 5.3\cdot10^{-5} -> 1.3\cdot10^{-8} | Improved |
Closest papers by meaning
- Improving Energy Natural Gradient Descent through Woodbury, Momentum, and Randomization (NeurIPS 2025 · similarity 0.93)
- Near-optimal Sketchy Natural Gradients for Physics-Informed Neural Networks (ICML 2025 · similarity 0.91)
- Dual Cone Gradient Descent for Training Physics-Informed Neural Networks (NeurIPS 2024 · similarity 0.88)
- Gradient Alignment in Physics-informed Neural Networks: A Second-Order Optimization Perspective (NeurIPS 2025 · similarity 0.88)
- ANaGRAM: A Natural Gradient Relative to Adapted Model for efficient PINNs learning (ICLR 2025 · similarity 0.88)
- Challenges in Training PINNs: A Loss Landscape Perspective (ICML 2024 · similarity 0.88)
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.
