A Simple Nonlinear Classifier Using a Multimode Optical Chip

Abstract

The photonic neural network is emerging as a promising technology for fast and power-efficient computing. So far, optics mainly perform linear transformations, and it remains challenging to integrate nonlinear activation function. The cover image shows an optical neural chip with only one multimode waveguide and a few electrodes. Though extremely simple, this chip constructs a modular network with both linear and nonlinear transformations. This work (see article number 2300253 by Ziyang Zhang and co-workers) offers an alternative route to exploiting the modulated multimode interference for photonic computing applications.