论文部分内容阅读
采用具有瞬态混沌特性的神经网络 ( TCNN)解任务分配问题 .该方法利用神经元的自反馈产生混沌动态 ,由于混沌动态特性具有很强的搜索全局最优解的能力 ,有效地避免了传统 Hopfield神经网络 ( HNN)极易陷入局部极小的缺陷 ;同时利用时变参数控制混沌行为 ,使网络在经过一个短暂的混沌倒分岔后逐渐趋于一般的 Hopfield神经网络 ,保证网络收敛到一个最优或近似最优的稳定平衡点 .仿真结果表明 ,TCNN解任务分配问题时 ,总能收敛到全局最优或几乎接近全局最优 ,同时具有更高的搜索效率 .另外 ,还用此方法求解了属于NP-完全问题的实时分布处理系统中的任务分配问题 .
This paper uses the neural network (TCNN) with transient chaos characteristics to solve the task assignment problem. This method uses the self-feedback of neurons to generate chaotic dynamics. Since the chaotic dynamics has a strong ability to search the global optimal solution, Hopfield neural network (HNN) can easily fall into local minimum defects. At the same time, the chaotic behavior is controlled by using time-varying parameters, which make the network converge to a general Hopfield neural network after a brief chaos inverse bifurcation, and ensure that the network converges to a The optimal or nearly optimal stable equilibrium point.The simulation results show that TCNN always converges to the global optimal or almost global optimal solution to the problem of task assignment and has higher search efficiency.In addition, Solve the task distribution problem in the real-time distributed processing system which belongs to the NP-complete problem.