基于Gibbs采样与概率分布估计的移动云数据存储  

Mobile Cloud Data Storage Based on Gibbs Sampling and Probability Distribution Estimation

作　　者：李又玲[1] 常致全[1] 

机构地区：[1]四川大学计算机学院,成都610065

出　　处：《计算机工程》2017年第1期13-19,共7页Computer Engineering

摘　　要：为提高移动云数据存储远程服务器的计算和存储能力,提出一种改进的移动云数据存储算法。利用表决数据分配和表决数据处理框架,构建考虑节点失效概率的重采样期望传播时间计算模型,并建立整合能源效率和容错性的表决动态网络。采用概率分布估计对动态网络模型进行存储路径优化,应用Gibbs采样解决分布估计的样本数据高维耦合和无监督训练问题。实验结果表明,与贪心算法、随机放置算法和分布估计算法相比,该算法具有更高的能源效率和存储可靠性。In order to improve the computing and storage capacity of mobile cloud data storage remote server,this paper proposes an improved mobile cloud data storage algorithm.Firstly,it constructs resampling expected propagation time calculation model by considering node failure probability with the voting data distribution and voting data processing framework,and establishes the dynamic voting network integrating energy efficiency and fault tolerance.It uses the probability distribution estimation method to optimize the storage routes of dynamic network model.At the same time,it uses Gibbs sampling to solve the problems of high-dimensional coupling and unsupervised training of sample data and non supervision training.Experimental results show that compared with the greedy algorithm,random placement algorithm and Estimation of Distribution Algorithms （EDAs）,the proposed algorithm has high energy efficiency and storage reliability.

关 键 词：GIBBS采样 分布估计 重采样 移动云 数据存储 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]