Laser delay dynamics for information processing
Held by Bogdan Penkovsky (University of Franche-Comté, Besançon,
13.07.2016, 12:15 Uhr
Motivation behind our research in dynamical systems is implementation of biologically inspired hardware. In this talk, we are going to discuss patterns known as chimera states which spontaneously emerge in delay dynamics. A variety of chimera configurations can coexist depending uniquely on initial conditions. These configurations can be regarded as memory stored directly in the dynamics.
Later in the talk, we will reuse the same dynamical system for the purposes of reservoir computing (RC). RC technique refers to a generalized machine learning approach for tasks of different complexity. Its strength is the possibility to deal with complex data such as images, sounds, time series and other 'fuzzy' logic where conventional computers' performance falls considerably or requires intricate algorithms to be developed. We will discuss how delay dynamics would allow us to achieve the principles of RC in hardware.