GPT-PINN and TGPT-PINN: Advanced Physics-Informed Neural Networks for Parametric PDEs

Explore the innovative GPT-PINN and TGPT-PINN approaches in this 59-minute talk by Dr. Yanlai Chen from UMass Dartmouth. Delve into the world of Physics-Informed Neural Networks (PINN) and discover how these new meta-learning paradigms address challenges in parametric PDEs. Learn about the unique structure of GPT-PINN as a network of networks, its adaptive learning capabilities, and its efficiency in generating surrogate solutions across parameter domains. Gain insights into the TGPT-PINN's nonlinear model reduction techniques through the addition of a transformation layer. Understand the potential applications and benefits of these advanced methods in numerical analysis and scientific computing, presented by an expert in the field of mathematics and machine learning algorithms.

DDPS | ‘GPT-PINN and TGPT-PINN