Recursive Estimation of State-Space Noise Covariance Matrix by Approximate Variational Bayes
Résumé
This working paper considers state-space models where the variance of the observation is known but the covariance matrix of the state process is unknown and potentially time-varying. We propose an adaptive algorithm to estimate jointly the state and the covariance matrix of the state process, relying on Variational Bayes and second-order Taylor approximations.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|