Speaker
Jingyi Tang
(Stanford University)
Description
We present a differentiable 1D free-electron laser (FEL) simulation framework based on automatic differentiation. In contrast to conventional FEL codes that provide only forward simulation, the differentiable formulation enables efficient evaluation of gradients of FEL observables with respect to beam, undulator, and model parameters. This makes it possible to apply gradient-based methods to optimization, sensitivity analysis, parameter fitting, and inverse problems within a physics-based FEL mode
| I have read and accept the Privacy Policy Statement | Yes |
|---|
Authors
Jingyi Tang
(Stanford University)
Zhirong Huang
(SLAC National Accelerator Laboratory)