Input compensation learning: Modelling dynamical systems
André Frank Krause, Volker Dürr, Thomas Schack, Holk Cruse · 2011
A special class of recurrent neural networks, Input Compensation (IC) networks, is applied to model two exemplary dynamical systems, the Van-der-Pol Oscillator and the Figure-Eight pattern. IC-learning results in compact networks that provide insights into the underlying properties of the modelled system.