Neural Differentiable Physics: Unifying Numerical PDEs and Deep Learning for Data-Augmented Computational Physics

Explore the cutting-edge field of Neural Differentiable Physics in this comprehensive 1-hour 28-minute talk by Dr. Jianxun Wang from the University of Notre Dame. Delve into the challenges faced by traditional numerical models in predictive modeling and simulation of complex physical systems. Discover how the integration of advanced machine learning techniques with physics principles is revolutionizing computational mechanics. Learn about the innovative SciML framework that unifies classic numerical PDE solvers and deep learning models, offering a novel approach to modeling complex physical systems. Gain insights into the integration of numerical PDE operators into neural architectures, enabling the fusion of prior knowledge, multi-resolution data, and deep neural networks through differentiable programming. Understand how this approach differs from existing frameworks like Physics-Informed Neural Networks (PINNs) and its potential to usher in a new era of understanding and modeling complex physical systems with far-reaching implications for science and engineering applications.