文档介绍：哈尔滨工业大学工学硕士学位论文摘要 Cm着通信网特别是计算机网络的飞速发展,路由选择技术成为影响网络性能的重要因素。究的热点。并且, 基于这原因,网络领域的路由问题己经成为近年来人们研自从Hopfield成功的将神经网络应用于解决旅行商问题之许多人将此作为一种优化方法用于各种优化本文首先介绍了神经网络的基本概念和算法绍了神经网络的优缺后点其次,本文针对路由问题与旅行商问题的相似性,提出了一种基于 Hopfield神经网络的虚令牌动态落蔽瑟憩一法;渗Hopfield神经网络,其神经元之间的连接强度固定不变,通信网络的物理链路值和呼叫请求被反映到神经元的偏置电流中,易于用硬件实现。乎然后,本文针对神经网络解决这类问题容易陷入局部最优解的缺点,提出了一种改进这种策略,虽然不一定能达到全局极小,但是能得到较好的解。并对一个10节点的网络进行了仿真,仿真结果表明,该策略可行。最后,验证了该算法的性能,即当通信网络中的业务量发生变化时,算法能够自动调整最短路径与均衡负载之间的关系与最短路算法和最小负载算法作了比较,计算仿更多的连接·即降低了网络的呼损率分损率作为性能指标, 表明,该算法能接入关键词路由选择算法;Hopfield神经网络;虚令牌;呼损率哈尔滨工业大学工学硕士学位论文 Abstract Routing has been an important factor that has a significant impact on work's performance with rapid development of the communication networks especially of the computer networks. For this reason the area of network routing has been the subject of intensive research for many years. Since the use of works to solve optimization problems was initiated by Hopfield, many people have used it to solve the various optimization problems. In this thesis, fundamental concepts and algorithm are proposed first, and the merits and demerits of neural networks second Secondly, a dynamic routing algorithm based on virtue token and Hopfield neural network is proposed according to the similarity of the routing problems and the traveling salesman problem. In the algorithm, the interconnection matrix of the neural network is fixed, and the link costs and call requests of work are mapped into the biases of the neurons. So the Hopfield neural network is easy to be implemented into the hardware Thirdly, an improvement method is proposed for the work in solving optimization problems often getting local optimal solutions. And simulate with a network of 10 nodes. Results show that the improvement is valid Ailast, the algorithm performance is validated. The algorithm can adjust the b