Academic Journal
Computing Systems
| العنوان: | Computing Systems |
|---|---|
| المؤلفون: | José A. Gámez, Juan L. Mateo, José M. Puerta |
| المساهمون: | The Pennsylvania State University CiteSeerX Archives |
| المصدر: | http://pgm08.cs.aau.dk/Papers/21_Paper.pdf. |
| المجموعة: | CiteSeerX |
| الوصف: | Dependency networks are a probabilistic graphical model that claim several advantages from other models like Bayesian networks and Markov networks, for instance. One of these advantages in general dependency networks, which are the object of study in this work, is the ease of learning from data. Nonetheless this easiness is also the cause of its main drawback: inconsistency. A dependency network cannot encode the probability distribution underlay in the data but an approximation. This approximation can be enough good for some applications but not in other cases. In this work we make a study of this inconsistency and propose a method to reduce it. From the conclusions we have taken from this analysis we have developed an algorithm that has to be run after the standard learning algorithm yields its solution. Our method is an heuristic approach so we cannot assure that the resulting model is fully consistent, however we have carried out some experiments which make us to think that it produces high quality models and therefore is advisable its use. |
| نوع الوثيقة: | text |
| وصف الملف: | application/pdf |
| اللغة: | English |
| Relation: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.321.49; http://pgm08.cs.aau.dk/Papers/21_Paper.pdf |
| الاتاحة: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.321.49 http://pgm08.cs.aau.dk/Papers/21_Paper.pdf |
| Rights: | Metadata may be used without restrictions as long as the oai identifier remains attached to it. |
| رقم الانضمام: | edsbas.CB9294B6 |
| قاعدة البيانات: | BASE |
| الوصف غير متاح. |