All-optical deep feed forward network based on nonlinear microresonators for telecom applications (Conference Presentation)

Recently, there have been a renewed interest on brain-inspired (neuromorphic) computing schemes that are directly implemented on a physical platform. In fact, such optical or electrical platforms allow performing important computational tasks at a speed much higher than the software-based counterparts. Here we propose to use neuromorphic silicon photonics to recover data integrity directly in the optical domain outperforming the electronic performances. We demonstrate a silicon photonic Feed Forward Network that can be trained to solve several tasks directly in the optical domain: linear and non-linear distortion recovering, demodulation of complex modulated data and error correction.