基于BD度量的简单贝叶斯优化学习算法

摘要

贝叶斯优化算法是利用贝叶斯网络匹配进化种群的优良解集而产生新的染色体来体现种群的进化.在贝叶斯网络对种群进行匹配的过程中,贝叶斯网络结构越复杂,种群的进化信息描述越完整,进化质量越高,但运算速度相对来说越慢；相反,贝叶斯网络越简单,算法描述的种群的进化信息越少,进化质量越差,但却能够提高算法的运算速度.基于此,给出了简单贝叶斯优化与复杂贝叶斯优化定义.针对简单贝叶斯网络提出了基于BD度量的三步结构学习算法,并给出了一个利用这种算法进行贝叶斯网络结构学习的例子.%Bayesian optimization algorithm is a method using Bayesian network matching optimal population product new population. In this process, the structure of Bayesian network is more complex, the evolutionary information of the population is more, and the quality of evolution is better, but the speed of evolution is more slowly. However, if the structure of Bayesian network is more simply, the evolutionary information of the population is less, and the quality of evolution is worse, but the speed of evolution is faster. This paper presents the conceptions of simple Bayesian optimization algorithm and complex Bayesian optimization algorithm based of this. And a new three-phase method from database to simple Bayesian network is given based on the conception. To evaluate this algorithm, an example in the third section is presented.