A Generalized Updating Rules Using Hopfield-Type Neural Networks for Optimization Problems
-
摘要: 基于离散Hopfield-型网络和延迟离散Hopfield-型网络求解优化问题提出了两种一般 演化规则,演化序列的动态阈值是这些规则的重要特征,并获得了收敛性定理.推广了已有的 离散Hopfield-型网络和延迟离散Hopfield-型网络的收敛性结果,给出了能量函数局部极大值 点与延迟离散Hopfield-型网络的稳定态的关系的充分必要条件.鉴于延迟离散Hopfield-型网 络更有效地应用于优化计算问题,给出了一般分解策略.实验表明与离散Hopfield-型网络的 算法相比,文中提出的算法既有较高的收敛率又缩短了演化时间
-
关键词:
- Hopfield-型网络 /
- 延迟 /
- 收敛性 /
- 稳定态
Abstract: This paper presents two generalized updating rules based on Hopfield-type neural networks (with delay or without delay) for optimization problems. These rules are characterized by dynamic thresholds of the updating sequence. Convergence theorems of discrete Hopfield-type neural networks with delay are obtained, which extend the exsiting convergence results. Also obtained is a sufficient and necessary condition for the relation between the stable states of neural networks and the points of local maximum value of energy function. Decomposed strategy is given in order to apply the Hopfield-type neural networks with delay to optimization problems effectively. Finally, the experimental results demonstrate that the given algorithm improves the convergence rate and decreases the updating time when compared with Hopfield-type neural network without delay.-
Key words:
- Discrete Hopfield-type neural network /
- delay /
- convergence /
- stable state
计量
- 文章访问数: 2361
- HTML全文浏览量: 89
- PDF下载量: 1342
- 被引次数: 0