Conditions for attaining the global minimum in maximum likelihood system identification
Research output: Contribution to journal › Article › peer-review
Electronic versions
DOI
Maximum likelihood estimation(MLE) is a popular technique in both open and closed loop identification. However when the landscape of likelihood function has several local minima, gradient based optimization might end up with local convergence. To avoid this, various non-local-minimum conditions are derived in this paper. Here we consider different model structures, in particular Output-Error, ARMAX, and Box-Jenkins models
| Original language | Unknown |
|---|---|
| Pages (from-to) | 1110-1115 |
| Number of pages | 6 |
| Journal | IFAC Proceedings Volumes (IFAC-PapersOnline) |
| Volume | 42 |
| Issue number | 10 |
| Early online date | 19 Feb 2010 |
| DOIs | |
| Publication status | Published - 21 Apr 2016 |
| Externally published | Yes |