Group Shuffled BP算法的密度演进和高斯近似  

Gaussian Approximation of Density Evolution Based on Group Shuffled BP Algorithm

作　　者：陈文[1,2] 刘斌彬[1,2] 白栋[1,2] 葛启宏[1,2] 

机构地区：[1]国家广电总局广播科学研究院 [2]北京泰美世纪科技有限公司,北京100097

出　　处：《通信技术》2010年第4期53-55,58,共4页Communications Technology

基　　金：国家科技支撑计划项目(No:2008BAH25B03)

摘　　要：为了分析Group Shuffled BP译码算法的收敛性能,同时简化密度演进计算的复杂度,在证明对称性条件的基础上,提出了基于Group Shuffled BP译码算法的密度演进的高斯近似。从而将密度演进中计算消息概率密度的无限维问题,简化为跟踪高斯分布均值的一维问题。仿真结果表明,该方法具有较高的精确度,可以有效分析Group Shuffled BP译码算法的收敛性能。In order to analyze the convergence performance of Group Shuffled Belief Propagation（BP） algorithm and reduce the computation complexity of density evolution,the Gaussian approximation of density evolution based on group shuffled BP algorithm is proposed.The infinite dimensional problem of calculating probability densities of messages is converted into one-dimensional problem of tracking the means of Gaussian distributions.Simulation results show that the method is of high accuracy,and could effectively analyze the convergence performance of Group Shuffled BP algorithm.

关 键 词：GROUP Shuffled BP译码算法 密度演进 高斯近似 

分 类 号：TN911.22[电子电信—通信与信息系统]