Optimizing dynamical states for high-speed photonic computing
Résumé
Photonic neuromorphic computing is the in-tersection of sub nanosecond fluctuations with the perfor-mance and energy efficiency of the human brain. Thesesystems operate based on a nonlinear transformation of in-put streams, and by fully understanding the attractor dy-namics of the transformation it is possible to optimize thesedynamics for a computing task. By optimizing the dynam-ical state of our neural net, we are able to process over 70million digits per second.