|
Descripción
|
|
|---|---|
| Supervised classification in mixed domains with probabilistic graphical models: we have adapted a set of algorithms taken from Bayesian multinomial networks to conditional Gaussian networks. We also have proposed novel classifier induction algorithms based on the particularities of conditional Gaussian networks. Moreover, we have proposed the novel kernel based Bayesian network paradigm which extends the idea of flexible naive Bayes breaking with the parametric assumptions. In addition, we have adapted some of the algorithms proposed for Bayesian multinomial networks to this novel paradigm. In order to present the kernel based Bayesian network paradigm, the mixed Gaussian kernel distribution is introduced | |
|
Internacional
|
Si |
|
ISBN
|
|
|
Tipo de Tesis
|
Doctoral |
|
Calificación
|
Sobresaliente cum laude |
|
Fecha
|
21/05/2010 |