基于新型二进制自适应差分演化算法的贝叶斯网络学习

Bayesian Network Learning Based on the New Binary System Adaptive Differential Evolution Algorithm

贝叶斯网络是处理不确定的专门知识和推理的最流行的方法,并广泛应用于大量研究领域。贝叶斯网络学习的主要策略是利用统计评分来选择最优网络的候选者。论文提出了基于新型二进制自适应差分演化算法的贝叶斯网络学习(BDENBAL),该方法采用一种自适应的0/1矩阵作为比例因子,并通过交叉和变异算子来实现贝叶斯网络的学习过程中的信息交换。然后根据贝叶斯信息标准(BIC)评分从网络空间中选择贝叶斯网络候选者。实验结果证明本文提出的方法具有优良的性能。

The Bayesian network is the most popular way to deal with uncertain knowledge and reasoning,and widely used inplenty of research fields. The strategy of Bayesian network learning is to choose the candidate of the optimal network by using the sta-tistical score. The Bayesian network learning is proposed based on the new binary system adaptive differential evolution algorithm（BDENBAL）. BDENBAL uses a self-adaptive 0/1 matrix as a proportional factor,And through the cross and mutation operator torealize the information exchange in the learning process of Bayesian network. Then,according to the Bayesian Information Criterion（BIC）scoring standard,the candidate is selected from the network space. The experimental results show that the proposed methodhas excellent performance.