Unlocking Nonlinear 850-nm VCSEL Dynamics for AI Computing in Radio-Over-Fiber Networks

Abstract

The paradigm of joint communication and computation (J2C) is proposed to integrate artificial intelligence (AI) features within mobile front-haul infrastructures that exploit analog radio-over-fiber (ARoF) technology. The nonlinearity of a vertical cavity surface emitting laser (VCSEL)-based ARoF system is used for processing the transmitted data. In a classical communication system, the data connected at one side of the network is transmitted through the network to the other side for being processed. In this letter, we discover that nonlinear properties of VCSEL enable the neural network computing when signals are transmitted in the network, drastically reducing the computing consumption.