A HFC-based Bayesian Network Structure Learning Algorithm
| Autor: | Ming Zhao, Qing Zhou, Guangrong Bian |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
education.field_of_study
Population Process (computing) Bayesian network 010103 numerical & computational mathematics 02 engineering and technology Mutual information Space (commercial competition) 01 natural sciences Class (biology) Local optimum Bayesian information criterion 020204 information systems 0202 electrical engineering electronic engineering information engineering 0101 mathematics education Algorithm |
| Zdroj: | 2017 International Conference on Industrial Informatics - Computing Technology, Intelligent Technology, Industrial Information Integration (ICIICII). |
| DOI: | 10.1109/iciicii.2017.11 |
| Popis: | For the problem that Bayesian network structure learning algorithm is easy to fall into local optimum, this paper proposes a scoring search method based on HFC (Hierarchical Fair Competition) model. It divides the population into different grades according to its fitness, and carries on the evolutionary method of intra class competition and inter level migration to ensure the diversity of the population in the process of evolution and avoid the search into the local optimum. At the same time, the mutual information theory and BIC (Bayesian Information Criterion) scoring criterion are used to calculate the initial network structure, reduce the search space and shorten the search time. Finally, the algorithm is used to study the network structure of the classical data sets——Asia, and the results show the decrease in the number of extra edges and missing edges which better describe the network structure contained in the data. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|