Neuromorphic Computing in Ginzburg-Landau Polariton-Lattice Systems

Reservoir computing is an increasingly popular approach to hardware implementations of neural networks. It does not require fine tuning of system parameters, and holds promise for high processing rates in photonic systems. The authors demonstrate how this concept can be applied in systems described by the complex Ginzburg-Landau equation, one of the fundamental models of wave phenomena. In particular, it is predicted that lattices of semiconductor microcavities could be used for information processing at data rates on the order of 1 Tbit/s, two orders of magnitude higher than the record to date in optical systems.