Stability and Bifurcation Analysis of Nonlinear PDEs via Random Projection-based PINNs: A Krylov-Arnoldi Approach
Not on the map. Why?
Abstract (arXiv)
We present a numerical framework for the stability and bifurcation analysis of nonlinear partial differential equations (PDEs) in which the solutions are sought in the function space spanned by physics-informed random projection neural networks (PI-RPNNs), and discretized via a collocation approach. These are shallow, single-hidden-layer networks with randomly sampled and fixed a priori hidden-layer weights; only the linear output layer weights are optimized, reducing training to a single least-squares solve. This linear output structure enables the explicit generalized eigenvalue formulation governing the linear stability of stationary solutions, separating the physical domain interior dynamics from the algebraic constraints imposed by boundary conditions (BCs), at no additional training cost and without requiring additional PDE solves. However, the random projection collocation matrix is inherently numerically rank-deficient, rendering naive eigenvalue computation unreliable and contaminating the true eigenvalue spectrum with spurious near-zero modes. To overcome this limitation, we introduce a matrix-free shift-invert Krylov-Arnoldi method that operates directly in weight space, avoiding explicit inversion of the numerically rank-deficient collocation matrix and enabling the reliable computation of several leading eigenpairs of the physical Jacobian - the discretized Frechet derivative of the PDE operator with respect to the solution field, whose eigenvalue spectrum determines linear stability. We further prove that the PI-RPNN-based generalized eigenvalue problem is almost surely regular, guaranteeing solvability with standard eigensolvers, and that the singular values of the random projection collocation matrix decay exponentially for analytic activation functions, rigorously explaining the observed ill-conditioning.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 33 of its 91 references that are in the map. First posted on arXiv 23 Mar 2026.
Builds on
33 of 91 references.
References · 33
- 2019DeepXDE: A deep learning library for solving differential equationscited by 619
- 2017Solving high-dimensional partial differential equations using deep learningcited by 556
- 2021Learning the solution operator of parametric partial differential equations with physics-informed DeepOnetscited by 540
- 2021Physics-Informed Neural Operator for Learning Partial Differential Equationscited by 527
- 2019An Energy Approach to the Solution of Partial Differential Equations in Computational Mechanics via Machine Learning: Concepts, Implementation and Applicationscited by 118
- 2022Physics-Informed Deep Neural Operator Networkscited by 113
- 2020Local Extreme Learning Machines and Domain Decomposition for Solving Linear and Nonlinear Partial Differential Equationscited by 81
- 2019Physics Informed Extreme Learning Machine (PIELM) -- A rapid method for the numerical solution of partial differential equationscited by 67
- 2020The gap between theory and practice in function approximation with deep neural networkscited by 54
- 2025Optimizing the Optimizer for Physics-Informed Neural Networks and Kolmogorov-Arnold Networkscited by 42
- 2023AI-Aristotle: A Physics-Informed framework for Systems Biology Gray-Box Identificationcited by 29
- 2021On Computing the Hyperparameter of Extreme Learning Machines: Algorithm and Application to Computational PDEs, and Comparison with Classical and High-Order Finite Elementscited by 26
21 more
- 2022Constructing coarse-scale bifurcation diagrams from spatio-temporal observations of microscopic simulations: A parsimonious machine learning approachcited by 25
- 2023Training Neural Networks is NP-Hard in Fixed Dimensioncited by 23
- 2024Score-Based Physics-Informed Neural Networks for High-Dimensional Fokker-Planck Equationscited by 22
- 2023Sampling weights of deep neural networkscited by 20
- 2022Parsimonious Physics-Informed Random Projection Neural Networks for Initial-Value Problems of ODEs and index-1 DAEscited by 19
- 2024RandONet: Shallow-Networks with Random Projections for learning linear and nonlinear operatorscited by 16
- 2021Learning normal form autoencoders for data-driven discovery of universal,parameter-dependent governing equationscited by 15
- 2025Simulating Three-dimensional Turbulence with Physics-informed Neural Networkscited by 14
- 2025Learning and discovering multiple solutions using physics-informed neural networks with random initialization and deep ensemblecited by 13
- 2024Random Projection Neural Networks of Best Approximation: Convergence theory and practical applicationscited by 11
- 2023Slow Invariant Manifolds of Singularly Perturbed Systems via Physics-Informed Machine Learningcited by 10
- 2025Enabling Local Neural Operators to perform Equation-Free System-Level Analysiscited by 10
- 2024Nonlinear Discrete-Time Observers with Physics-Informed Neural Networkscited by 8
- 2024Koopman-informed recurrent neural networkscited by 8
- 2024Gradient-Free Training of Recurrent Neural Networks using Random Perturbationscited by 6
- 2024Efficient Training of Deep Neural Operator Networks via Randomized Samplingcited by 6
- 2024Neural networks for bifurcation and linear stability analysis of steady states in partial differential equationscited by 5
- 2023Discrete-Time Nonlinear Feedback Linearization via Physics-Informed Machine Learningcited by 4
- 2025Physics-informed neural networks for high-dimensional solutions and snaking bifurcations in nonlinear latticescited by 3
- 2025Equation-Free Coarse Control of Distributed Parameter Systems via Local Neural Operatorscited by 1
- 2025HEATNETs: Explainable Random Feature Neural Networks for High-Dimensional Parabolic PDEscited by 1
Built on by
Not cited yet.