Silicon microring synapses enable photonic deep learning beyond 9-bit precision

Deep neural networks (DNN) consist of layers neurons interconnected by synaptic weights. A high bit-precision in weights is generally required to guarantee accuracy many applications. Minimizing error accumulation between also essential when building large-scale networks. Recent demonstrations photonic are limited due crosstalk and the sensitivity optical components (e.g., resonators). Here, we experimentally demonstrate a record-high precision 9 bits with dithering control scheme for synapses. We then numerically simulated impact increased on wireless signal classification application. This work could help realize potential practical, real-world tasks.