1. Physically transparent diffractive optical networks
- Author
-
Wu, Ruitao, Fang, Juncheng, Pan, Rui, Lin, Rongyi, Li, Kaiyuan, Lei, Ting, Du, Luping, and Yuan, Xiaocong
- Subjects
Physics - Optics - Abstract
Inspired by neural network algorithms in deep learning, diffractive optical networks have arisen as new platforms for manipulating light-matter interactions. Seemly inherited from the deep learning black box nature, clear physical meanings have never been given for these complex diffractive networks at the layer level, even though the systems are visible in physical space. Using exemplified mode conversion systems, this work showed how various physical transformation rules within diffractive networks can be unveiled with properly defined input/output relations. Surprising physical transformation division phenomenon and an optical analogy of gradient-vanishing-effect have been observed and discussed for high-dimensional mode sorting tasks. The use of physical transparency for the effective design of parameter-varying networks is also demonstrated. These physically transparent optical networks resolve the contradiction between rigorous physical theorem and operationally vague network structure, and pave the way for transparentizing other physical neural networks.
- Published
- 2024