Study of the effective number of parameters in nonlinear identification benchmarks

This poster discusses the importance of the notion of effective number of parameters as a measure of model complexity. Exploiting this concept allows a fair comparison of models obtained from different model classes. Several illustrative examples of linear and nonlinear models are presented to provide more insight in the problem. A number of possibilities to reduce the model complexity are also discussed, including regularization techniques and an alternative approach based on rank reduced estimation. These ideas are then applied to two nonlinear real world problems: the Wiener-Hammerstein and the Silverbox benchmarks.