Fudan University Develops AI to Reconstruct Complex Physics Systems from Limited Data
-
Fudan University researchers developed a machine learning framework called Hamiltonian Neural Koopman Operator (HNKO) to predict complex Hamiltonian systems.
-
HNKO integrates mathematical physics principles to reconstruct high-dimensional Hamiltonian systems from limited, noisy data.
-
Key capability is discovering new conservation laws solely from observational data, enabling accurate prediction despite noise.
-
Demonstrated on complex physical models with hundreds to thousands of degrees of freedom, showing promise for understanding intricate dynamical systems.
-
Signifies critical advancement in leveraging AI to advance fundamental understanding of physics and mathematics.