Reservoir Computing With A Spatially Extended Network Of Semiconductor Lasers

Classical information processing based on the von Neumann concept have their limitations in solving complex problems like recognition of objects, speech etc. An alternative information processing approach is machine learning. Within the large field of machine learning, Reservoir Computing is a novel concept which is well suited for hardware realizations.
Inspired by the human brain information is processed in a nonlinear network. In my talk I will introduce a concept for realizing such a network with semiconductor lasers. For this we use a chip with 63 lasers, which we intend to couple with a diffractive optical element. I will introduce the theoretical framework, ray-tracing simulations of the system and its initial characterization.