Self organized classification of chaotic domains from a nonlinear attractor

We propose a method to use self organizing neural networks to extract information out of nonlinear dynamic systems for control. Nonlinear strange attractors are educed by these systems or the attractors can be reconstructed. These attractors are partitioned by a newly developed self organizing neural network. Thus the stream of system states is transformed into a stream of symbols, which can now serve as a basis for further investigation or control. We believe that controlling and understanding such nonlinear or chaotic systems is easier, when using the information within the stream of extracted symbols