A Photonic Implementation of Reservoir Computing
Kristof Vandoorne, Peter Bienstman · Ghent University Academic Bibliography (Ghent University) · 2007
Reservoir Computing [1] is a new approach to study and use Neural Networks, which try to mimic a brain-like intelligence. It uses memory and feedback in the reservoir to address more complex classification and recognition tasks as extracting time correlated features in problems like speech processing [2]. This has already been realized using software implementations but a hardware approach needs yet to be realized. Photonics offers a good platform to achieve this in a fast and economic way. The main idea is to build a photonic reservoir by exploiting and combining fast non linear light interactions to manipulate light.