Neuromorphic Computing in Ginzburg-Landau Polariton-Lattice Systems

The availability of large amounts of data and the necessity of processing it efficiently have led to the rapid development of machine-learning techniques. To name a few examples, artificial-neural-network architectures are commonly used for financial forecasting, speech and image recognition, robotics, medicine, and even research. Direct hardware for neural networks is highly sought for overcoming the von Neumann bottleneck of software implementations. Reservoir computing (RC) is a recent and increasingly popular bioinspired computing scheme that holds promise for efficient temporal information processing. We demonstrate the applicability and performance of RC in a general complex Ginzburg-Landau lattice model, which adequately describes the dynamics of a wide class of systems, including coherent photonic devices. In particular, we propose that the concept can be readily applied in exciton-polariton lattices, which are characterized by unprecedented photonic nonlinearity, opening the way to signal processing at rates of the order of 1 Tbit s(-1).