Author: Zeng, L.
Paper Title Page
MOP2WA03
The Feasibility of Neuron Network-based Beam-based Alignment  
 
  • L. Zeng
    SINAP, Shanghai, People's Republic of China
 
  Artificial neuron networks which inspired by biological neural networks have been widely used in various domains, including computer vision, machine translation, pattern/speech recognition, medical diagnosis and so on, due to its overwhelming superiorities. But it's not until recently that intelligent¬†algorithms have been introduced in light source field. M.P. Ehrlichman, Yi Jiao, Juhao Wu and A. Sanchez-Gonzalez did some work in this respect and got commendable results. Considering Shanghai X-ray Free-Electron Laser (SXFEL) conditions, we are urgent to improve the FEL performance, and fundamental technique turns out to be beam-based alignment. But it's difficult to implement this means in SXFEL due to the low electron beam energy resulting in uncontrollable orbit disturbance. Thus, a new method which is suitable for SXFEL is an eager desire. Here, we discuss the feasibility of neuron network-based beam-based alignment, and try to take it into reality in SXFEL. In fact, Hornik have proved, as early as 1989, that a single hidden layer feedforward networks can approximate any measurable function arbitrarily well, which provides the theoretical evidence to our suggestion.  
slides icon Slides MOP2WA03 [5.439 MB]  
Export • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)