Microrings as programmable temporal kernels enabling photonic AI beyond 100 Gbaud
This work redefine MRRs as programmable temporal convolution kernels by exploiting their impulse responses, enabling computation beyond the resonance linewidth, and realizes a multi-channel MRR convolution engine with 120.8 TOPS throughput and reduce the hardware complexity.
Shao-Jie Liu, Teng-Ji Xu, Ben-Shan Wang et al.
· Nature Communications · 0 citations