Optoelectronic Demonstration of the Nonlinear Activation Module for Optical Neural Networks

Abstract

In this letter, we experimentally demonstrated the viability of the optoelectronic nonlinear activation module in optical neural networks. By adopting the beam splitter, photo-diode, electronic processing unit and tunable attenuator, nonlinear activation function ReLU which is widely used in the state of the art neural networks can be implemented in an optoelectronic way. Compared with other optical and optoelectronic implementations of nonlinear activation function, this method has advantages in expanses and complexity. This method also has potential to be applied into multiple types of optical neural networks to further increase the prediction accuracy.