中文摘要：

给出了模糊网络期望最短路径问题的定义,提出一种并行模糊神经网络最短路径（PFNNSP）算法解决模糊网络最短路径问题。PFNNSP算法通过模糊模拟对网络中的边权进行估计,脉冲波在神经元之间的并行传播,相互激活搜寻任意一对节点之间的最短路径,算法回溯输出路径表示和路径长度。在随机生成的小规模数据集上的仿真实验表明,PFNNSP算法在边权服从三角模糊分布的网络中执行时间优于Dijkstra算法,在大规模路网信息数据集上的仿真实验表明,PFNNSP算法能够有效求解网络中的最短路径,并且算法在迭代次数和收敛速度上要优于Dijkstra算法和A＊搜索算法。

英文摘要：

This paper proposed a parallel fuzzy neural network shortest path （PFNNSP） algorithm to solve the expected shor- test path problem on fuzzy network. First, it defined the definition of fuzzy network＇ s expected shortest path problem. Next, PFNNSP algorithm combined fuzzy simulation was developed to estimate edges＇ length. In the PFNNP, the pulse wave spread in parallel between neurons and searched shortest path of any pair nodes, while the shortest path and path length were obtained with the use of backtrack. Experiments on datasets with different scales show that the proposed PFNNSP algorithm leads to shorter computing time when compared with the well-known Dijkstra algorithm and A ＊ algorithm.