论文部分内容阅读
实时优化求解快速包交换问题 (FPS)是提高网络性能的重要手段 .基于梯度下降法等数学规划方法 ,不能并行地实时地优化求解FPS问题 ,而基于Hopfield型神经网络和细胞神经网络的优化方法中 ,都只有单一粒度的细胞动力学方程和单一粒度细胞之间的相互作用 ,不仅收敛到平衡点的过程长 ,而且神经网络参数的选择和修正十分困难 .该文提出一种新的具有多粒度宏细胞的广义细胞自动机模型和方法 ,广义细胞自动机中的小粒度宏细胞聚合成可以独立演化的大粒度宏细胞 ,通过多粒度群体的不同程度群体智能的相互作用 ,能够比目前其他方法更快更有效地分布并行地优化求解FPS问题和其它类似的复杂的网络优化问题 .
Real-time optimization of fast packet exchange (FPS) is an important means to improve network performance.Based on the gradient descent method and other mathematical programming methods, FPS can not be optimized in real time in parallel, but based on Hopfield neural network and cellular neural network optimization methods , There is only a single granularity of the kinetic equation of the cell and the interaction between the single granular cells, not only the process of converging to the equilibrium point is long, but also the selection and modification of the neural network parameters are very difficult.This paper presents a new multi- Granular macrophages generalized cellular automata model and method, the small-scale macrophages in the generalized cell automata polymerize into large-scale macrophages that can evolve independently. Through the interaction of different degrees of group intelligence in the multi-granularity groups, Method to solve FPS problems and other similar complex network optimization problems faster and more efficiently and in parallel.