Block-Sparse Gauss-Newton Optimization of Finite Basis Physics-Informed Neural Networks: Sparse Automatic Differentiation Approach
math.NA, cs.LG, cs.NA
Submitted: 2026-11-09
Updated: 2026-11-09
Code: https://github.com/ax-ml/jax
Terminology
Sources
- Challenges in Training PINNs: A Loss Landscape Perspective
- Adam: A Method for Stochastic Optimization
- Multi-Preconditioned LBFGS for Training Finite-Basis PINNs
- Finite basis physics-informed neural networks with hard constraints for viscous fluid flow in highly perforated domains
- An Optimisation Framework for the Well-Conditioned Training of Physics-Informed Neural Networks
- Tensor-Train Compressed Separable PINNs: A Curvature-Aware Optimization Framework for Parametric PDEs in High Dimensions
- Scaling Nonlinear Optimization: Many Problems One GPU
- asdex: Automatic Sparse Differentiation in JAX
Related papers
- Do physics-informed neural networks (PINNs) need to be deep? Shallow PINNs using the Levenberg-Marquardt algorithm
- A Neural-preconditioned Poisson Solver for Mixed Dirichlet and Neumann Boundary Conditions
- Advanced Linear Algebra with Applications - Part I (Numerical linear algebra for PDEs, machine learning, and data assimilation)
- Operator learning meets inverse problems: A probabilistic perspective
- Second-order consistency for learning chaotic dynamics via randomized Jacobian matching
- Windowed thinning and query complexity for the bouncy particle and Zigzag samplers