Structure-Preserving Learning of High-Dimensional Lagrangian and Hamiltonian Systems

Explore the cutting-edge approaches for learning structure-preserving reduced-order models in high-dimensional Lagrangian and Hamiltonian systems during this one-hour talk by Boris Kramer from UC San Diego. Delve into the importance of preserving physically interpretable quantities like momentum, energy, and vorticity in numerical simulations across diverse fields such as structural mechanics, aerospace engineering, wave propagation, and soft robotics. Discover how incorporating Lagrangian and Hamiltonian structures into the learning framework can lead to stable, robust models with exceptional long-term predictive accuracy using fewer training data than non-structured methods. Gain insights into recent developments in this field and their applications, showcasing the potential for more efficient and accurate simulations of complex dynamical systems.

DDPS | ‘Structure-Preserving Learning of High-Dimensional Lagrangian and Hamiltonian Systems’