2507.06342

Total: 1

#1 SymFlux: deep symbolic regression of Hamiltonian vector fields [PDF] [Copy] [Kimi] [REL]

Authors: M. A. Evangelista-Alvarado, P. Suárez-Serrato

We present SymFlux, a novel deep learning framework that performs symbolic regression to identify Hamiltonian functions from their corresponding vector fields on the standard symplectic plane. SymFlux models utilize hybrid CNN-LSTM architectures to learn and output the symbolic mathematical expression of the underlying Hamiltonian. Training and validation are conducted on newly developed datasets of Hamiltonian vector fields, a key contribution of this work. Our results demonstrate the model's effectiveness in accurately recovering these symbolic expressions, advancing automated discovery in Hamiltonian mechanics.

Subjects: Machine Learning , Artificial Intelligence , Dynamical Systems , Symplectic Geometry

Publish: 2025-07-08 19:07:16 UTC