Over the last century mankind has been experiencing a tremendous growth of computational capabilities thanks to inventions like the transistor, which inclined us on the use of electrical components for information processing. But now we are facing physical limitations to further extend computer architectures as we know them. Optics offers some attractive properties to overcome some of these limitations. Implementing novel information processing techniques like machine learning allow optical computers to achieve remarkable computing capabilities even at tasks traditional computers struggle with.
In this seminar, some of the activities on optical computing in IFISC’s Nonlinear Photonics Lab will be presented. We will introduce the experimental scheme of all-optical reservoir computing and focus on two key properties: dynamical consistency of the response to external modulation, and memory as the system’s ability to remember injected information. We study these properties in dependence on the optical frequency detuning between the injected information and the light in the reservoir, and show that by changing this quantity we can control consistency and memory.