Scalable Neuromorphic Photonic Circuits

该项目旨在开发小规模硅光子神经形态（神经元和大脑启发）电路，其架构和使能技术允许此类电路在尺寸上进行缩放。有了足够数量的节点（光子神经元），这样的电路将能够在人工智能（AI）硬件中的非线性优化、二次规划和深度学习加速等方面实现实际应用。在这个项目中要特别解决的最大挑战是光子神经元的片上增益和控制。在项目过程中，将有多个光子芯片的流片（芯片制造），包括越来越复杂的神经形态电路，光学放大器芯片和CMOS控制电子器件。

The project aims to develop small-scale silicon photonic neuromorphic (neuron and brain-inspired) circuits with architecture and enabling technologies allowing such circuits to be scaled in size. With a sufficient number of nodes (photonic neurons), such circuits would enable practical applications in nonlinear optimization, quadratic programming, and deep learning acceleration in artificial intelligence (AI) hardware, to name a few. The top challenges to be addressed specifically in this project are on-chip gain and control of photonic neurons. In the course of the project, there will be multiple tape-outs (chip fabrications) of photonic chips comprising neuromorphic circuits of increasing complexity, optical amplifier chips, and CMOS control electronics.