"Aprendizaje estructural de redes bayesianas en entornos locales a través de una malla variable"

This research paper presents a proposal to incorporate local search into the model of Optimization based on Variable Meshes (VMO). In this proposal, three algorithms of search will be applied in Variable Environment (VNS) which will be in charge of exploiting the close regions to good solutions already found in order to improve the search process and to obtain better solutions that only using the VMO algorithm for structural training of Bayesian Networks. Once selected the hybridization of the VMO with local search to better results were obtained with the parameters, intensity and metric set out later in this project, will be compared with the results of other algorithms that are found in the article "A hybrid method for learning bayesian networks based on ant colony optimization". For the development and execution of the VMO algorithm with local search, were used some of the main classes of software "Environment for the development of graphical models probabilistic (Elvira)", which was selected because of its ability for the construction of Bayesian Networks and be an open source tool. Keywords: Bayesian Networks, Optimization based on Variable Meshes (VMO), Variable Neighbourhood Search (VNS).

• Autore principale: Pacheco Zamora, Irina Monserrat (author)
• Natura: bachelorThesis
• Lingua: spa
• Pubblicazione: 2015
• Soggetti: Redes, Bayesianas, Optimización; Basado en mallas variables; Búsqueda en entorno variable VNS
• Accesso online: http://repositorio.uteq.edu.ec/handle/43000/4052