Sebastian Alveteg, doctorant à la Chaire Photonique de CentraleSupélec et au laboratoire LMOPS, donnera une conférence intitulée « Reservoir computing with a photorefractive crystal« .
La conférence se déroulera le jeudi 20 mars à 13h00 dans la Salle A007 de CentraleSupélec.
Abstract:
Reservoir computing is a machine learning framework that utilises a reservoir with constant weights as the network. The constant nature of the reservoir allows the use of physical systems in its place. Many implementations exist of this physical reservoir computer, and we have experimentally investigated the use of a photorefractive crystal as a reservoir. When a photorefractive crystal is exposed to light, it experiences an index change due to the electro-optic effect working in tandem with carrier generation, transport and following recombination. We investigate the relation between the index modulation observed in the crystal and the performance of some typical reservoir computing tasks. Three tasks with different ideal reservoir dynamics are used to evaluate the reservoir at different points in the parameter space, giving insight into the nonlinearity and memory of the system.