东南大学学报（自然科学版）2012，Vol.42Issue(z1)：82-86,5.DOI:10.3969/j.issn.1001-0505.2012.S1.018
K2与模拟退火相结合的贝叶斯网络结构学习
Bayesian network structure learning combining K2 with simulated annealing
摘要
Abstract
Aiming at the problem that the Bayesian network structure learning algorithm based on simulated annealing usually cannot obtain the optimal network structure due to the fact that the ability of the model perturbation to pass through all over the model space is poor, an improved Bayesian network structure learning algorithm, K2SA, is presented combining K2 with simulated annealing. With the new nodes order generated by exchanging two nodes selected randomly among the present nodes order, the K2SA uses the K2 algorithm to learn a Bayesian network as a new state in the simulated annealing algorithm in order to improve the ability of the model to perturb globally. The algorithm records the optimal network structure obtained in the course of simulated annealing, which is optimized by the hill-climbing algorithm after the course of simulated annealing. The simulation results show that under the condition of sufficient samples, the K2SA can obtain an approximately globally optimal network structure, but its efficiency is a little poor.关键词
模拟退火/K2算法/模型扰动/贝叶斯网络/结构学习/节点序Key words
simulated annealing/ K2 algorithm/ model perturbation/ Bayesian networks/ structure learning/ nodes order分类
信息技术与安全科学引用本文复制引用
金焱,胡云安,张瑾,黄隽..K2与模拟退火相结合的贝叶斯网络结构学习[J].东南大学学报（自然科学版）,2012,42(z1):82-86,5.基金项目
军队科研基金资助项目. （）
