【作者】 汪鸣鑫；

【导师】 周绍梅；

【作者基本信息】 南昌大学 ， 计算机软件与理论， 2007， 硕士

【摘要】 本论文主要探讨人工神经网络Hopfield在最优化领域，特别是组合优化领域的应用。由于神经网络模型具有高度并行性，易于电路实现等特点，使得它越来越受到重视。虽然Hopfield网络在求解组合优化问题时得出惊人的结果，但同时还存在很多有待研究和改进的问题。特别是Hopfield网络在解决约束组合优化时，满足约束条件和得到高质量的解是需要平衡的一对矛盾，这使得网络容易收敛到能量函数的局部极小点，而非问题的全局最优解。另外，网络优化在很大程度上依赖于网络的参数，即参数鲁棒性较差。针对以上问题，本文提出了Hopfield网络的若干改进措施，而且将改进的Hopfield网络应用在一类指派问题，其中一个是将改进的Hopfield网络应用在非平衡B指派问题和文件分配问题上；另外一个是将其应用在Job-Shop车间调度问题上。更多还原

【Abstract】 This dissertation aims at chaotic artificial neural network for solving combinational optimization problems. Because the artificical neural network that realized easily by electric circuit is high concurrent,more and more people research it.Although the Hopfield network to solve combinatorial optimization problems came to the startling results But there is also much room for research and improvement.In particular,the Hopfield network to solve combinational optimization constraint,meet the constraint conditions and access to high-quality solution is a need to balance contradictory.Network converges to the energy function of local minima, rather than the global optimal solution.Network Optimization largely depends on network parameters that are less robust parameters.Hopfield network itself is a nonlinear dynamical system to solve nonlinear control problems unique advantages,but this side and the rarely studied.To solve the above problem,this paper presents a Hopfield network of corrective measures,and will improve the Hopfield network application in a class assignment: Hopfield network in the non-equilibrium Bassignment and document distribution applications; Based on the Hopfield networkJob Shop Scheduling application.更多还原