置信规则库参数学习的并行差分进化算法

为解决置信规则库中现有参数学习方法主要是串行算法且不适用于求解大数据下参数优化模型的问题,结合群智能算法中的差分进化算法和集群系统中分布式方法,提出了基于消息传递接口的并行参数学习方法。以输油管道检漏问题为例,对比分析了本算法与现有参数学习方法在收敛时的误差,并在不同结点数的集群系统中分析了本算法的加速比和效率。实验结果表明,并行的参数学习方法是有效可行的。

Parallel differential evolution algorithm for parameter learning of belief rule base

To solve the problem of the existing parameter learning approaches for Belief Rule Base (BRB) were mainly serial algorithms, and those approaches were unsuitable for handling parameter optimization model under the big data. The differential evolution algorithm of swarm intelligence algorithms and the distributed method of cluster systems were introduced to the BRB, and then a parallel parameter learning approach using message passing interface was proposed. A numeric example of the pipeline leak detection problem was given. The new approach was compared with the existing parameter approaches in terms of the convergence error, the speedup ratio and the efficiency of parallel algorithm with different nodes of the cluster system. The experimental results showed that the approach was feasibilitiness and effectiveness.