A scalable on-chip optical neural network chip based on partially coherent light sources

Abstract

The application discloses a scalable optical neural network chip on chip based on a partial coherent light source and belongs to the optical computing field.The optical neural network chip comprises a partial coherent light source, an input coding unit, a linear computing unit, a nonlinear activation unit, an output detection unit and a computing and control unit.The application replaces a narrow linewidth laser with the partial coherent light source to reduce the dependence of the system on coherent detection and improve the scaling capacity of optical matrix computation.The optical-electric-optical (O-E-O) nonlinear activation unit is adopted to realize the scalability of the network level and effectively improve the computing efficiency and energy utilization rate in combination with an optimized convolution computing architecture.The chip of the application is manufactured by using an on-chip integrated manufacturing process, can realize end-to-end optical inference, supports complex pattern recognition tasks and has the characteristics of high bandwidth, low energy consumption, scalability and the like and is suitable for large-scale artificial intelligence computation.

Images