Loading...
Please wait, while we are loading the content...
Similar Documents
Apprentissage du nombre d'états d'une chaîne de Markov cachée pour la reconnaissance d'images
| Content Provider | Semantic Scholar |
|---|---|
| Author | Brouard, Thierry Slimane, Mohamed Venturini, Gilles |
| Copyright Year | 1997 |
| Abstract | Learning images with hidden Markov models (HMM) is a difficult problem, especially because of the important volume of data to be learned by the HMM. Several algorithms exist for this purpose. This paper presents a hybrid algorithm for learning HMM. This algorithm optimizes both at the same time the number of states and the parameters (probabilities) of HMMs. It relies on a genetic search of a good model among an heterogeneous population of HMMs (with different architectures) and on a local optimization with a gradient algorithm (Baum-Welch). This new algorithm converges rapidly and increases significantly the learning of images and the accuracy of recognition compared to standard algorithms. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://documents.irevues.inist.fr/bitstream/handle/2042/12744/211_049.pdf?sequence=1 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
