Presentations
Stabilized Lagrangian Neural Networks for learning relativistic geodesics and chaotic dynamics from data.
Jointly learning Lagrangians and physical constraints from trajectory data with a Constrained Lagrangian Neural Network.
We will study Lagrangian Neural Networks, a method that allows us to effectively "learn" the Lagrangian function, a key theoretical quantity in classical mechanics, from observed data alone. And with the help of symbolic regression, we get the underlying mathematical expressions of the Lagrangian. This seminar invites you to consider the potential of AI in deriving theoretical laws from experimental data, showing how machine learning is increasingly integral to advancing physics.