Deep Neural Networks as Discrete Dynamical Systems: Implications for Physics-Informed Learning
A new research paper proposes viewing deep neural networks (DNNs) as discrete dynamical systems, drawing parallels to neural integral equations and their PDE forms. The study compares numerical solutions of Burgers' and Eikonal equations with those from Physics-Informed Neural Networks (PINNs), suggesting PINNs offer a distinct computational path. While PINNs may use more parameters and be less interpretable than traditional methods, their flexibility could be advantageous in high-dimensional problems where grid-based approaches fail. AI
IMPACT Proposes a new theoretical framework for understanding DNNs, potentially influencing future research in physics-informed machine learning.