|
| 摘要: |
| 本文主要利用Gauss 信念传播算法求解具有多个子系统的线性系统, 并结合深度优先遍历算法来减少Gauss 信念传播算法在这一类线性系统中的计算复杂度. 同时根据以往 Gauss 信念传播算法的收敛性结果来分析我们的算法在不同子系统的收敛情况.最后本文选取了几个具有挑战性的线性系统——这些线性系统都具有多个独立的子系统, 并将我们的算法对比以往的传统迭代法以及传统的Gauss信念传播算法,展示这些算法在具有多个子系统的线性系统之中的性质. |
| 关键词: Gauss 信念传播算法, 深度优先遍历,子系统,弱连通分量 |
| DOI： |
| 分类号:O241.4 |
| 基金项目: |
|
| Research on Optimization of GaBP Algorithm Based on DFS |
|
qiudezhi,chensiyuan
|
| Abstract: |
| This paper mainly uses Gauss Belief Propagation to solve linear systems with multiple subsystems, and combines the Depth-First Search to reduce the computational complexity of the Gauss Belief Propagation in this type of linear system. At the same time, based on the convergence results of previous Gauss Belief Propagation, we analyze the convergence of our algorithm in different subsystems. Finally, this paper selects several challenging linear systems which have multiple independent subsystems, and compares our algorithms with traditional iterative methods and traditional Gauss Belief Propagation to show the properties of these algorithms in linear systems with multiple subsystems. |
| Key words: Gaussian Belief Propagation,Depth-First Search,Subsystem,Weak connected components |
