| Autor: |
Pertti M. Mäkilä |
| Rok vydání: |
2003 |
| Předmět: |
|
| Zdroj: |
IFAC Proceedings Volumes. 36:1867-1872 |
| ISSN: |
1474-6670 |
| DOI: |
10.1016/s1474-6670(17)35032-2 |
| Popis: |
An optimal squared error based approximation problem for static polynomial models is solved. This problem is similar to an optiInal approximation problem for linear time-invariant (LTI) models. Corresponding absolute error based approximation problems are also studied. Model quality estimation is typically based on sample variance analysis of the squared error criterion. Error squaring, however, results in increased sample variability especially for error and noise distributions with heavy tails. Error analysis based on the use of the sum of the absolute values of the errors has advantages in such situations. Two model quality estimation methods for static polynomial models are suggested based on similar techniques for LTI models. These techniques extend to many other linear-in-parameters regression problems. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
