Self-Organizing Networks for Dynamic Systems

摘要

Self-organizing networks have been applied to modeling and control of dynamic systems in a relatively minor scale. There are two prototypical methods of applying self-organization in systems modeling, the time domain approach and the operator domain approach. In time domain, process states are directly mapped in the net, while in operator domain, the parameters describing the process behavior are mapped. Both methods are useful in different applications, and they deserve to be analyzed separately. In this publication, theoretical results are summarized that justify the usage of the neural network methodology in modeling of dynamic systems. New efficient adaptation algorithms are presented that can be used not only for this application field.