Electronic Challenges to the Development of Efficient Neuromorphic Photonic Processors

The last decade has seen a significant surge in applications related to artificial intelligence and machine learning. Neural networks are becoming increasingly complex, a complexity which can be matched by specially designed photonic hardware that can outperform electronic counterparts in latency and power. However, considerable electronic circuitry is required to control the photonic system. The paper's objective is to numerically estimate how the electronic system affects key performance metrics in neuromorphic photonic processors, including weight resolution, latency, and power consumption.