Competitive Dynamics and Self-Organization in Photorefractive Systems

摘要

Information can be manipulated and processed by a nonlinear dynamical system.This work investigates the properties of such nonlinear optical systems based upon real-time holographic materials known as photorefractive media. Our goal is to understand how and when a dynamical system can self-organize, or learn on its own to process information in a useful way. We have demonstrated systems, for example, that can separate multiple fiber optic communication channels or learn to classify incoming images. We have discovered that such self-organizing systems will exhibit a phase transition with the information content of the input as the critical parameter. The phase transitions are analogous to those common to thermodynamic systems. In order to better provide an analytical basis of treatment of our systems we have developed various analytical tools, in particular stability analysis tools. We have also looked at the photorefractive version of well known nonlinear dynamical problems such as the formation of hexagonal patterns in pumped photorefractive media. Nonlinear optics, Nonlinear dynamics, Photorefractive materials, Neural networks, Holography.