Aprendizaje estructural de redes bayesianas: Un enfoque basado en puntaje y búsqueda

One of the most recent knowledge representations under uncertainty are Bayesian Networks whose main captivation is the property to obtain such a representation from a large amount of data. The issue is that getting a network structure is a NP-hard problem –commonly a learning process–, so there has been a lot of learning work where one of the best known methods is called based scoring and search approach. This paper reviews the basic definition of Bayesian networks, the scoring-and- search-based approach and by-products, that is, the hybrid approach and the search for equivalence classes; in addition, describes some algorithms for each approach and gives a summary of results of recent work.