Learning Bayesian networks skeleton: a comparison between TPDA and PMMS algorithm
"Learning the bayesian network structure from a database is an NP-Hard problem for which the existent learning algorithms generally have exponential complexity. During this work in the Master, I did a bibliographic research as well as a comparison between two recent algorithms called TPDA and P...
Guardado en:
| Autor principal: | |
|---|---|
| Otros Autores: | |
| Formato: | Proyecto final de Grado |
| Lenguaje: | Español |
| Publicado: |
2020
|
| Materias: | |
| Acceso en línea: | http://ri.itba.edu.ar/handle/123456789/3098 |
| Aporte de: |
| Sumario: | "Learning the bayesian network structure from a database is an NP-Hard problem for which the existent learning algorithms generally have exponential complexity. During this work in the Master, I did a bibliographic research as well as a comparison between two recent algorithms called TPDA and PMMS (2005) that learns the skeleton of bayesian networks from data. These algorithms have the advantage of having polynomial complexity, and provide good results for learning. After having done a theoretical analysis of the algorithms, I continue with an empiric analysis that consisted in testing these algorithms on data generated from networks knew by the scientific community (I used ASIA and ALARM networks). These tests have been made with the help of the toolboxes developed in Matlab (FullBNT, BNT – SLP and CausalExplorer). The results I have gotten by this analysis have permitted me to make some interesting conclusions about the efficiency and the limits of application of these algorithms." |
|---|