RW-MC:基于随机游走的自适应矩阵填充算法  

RW-MC: self-adaptive random walk based matrix completion algorithm

作　　者：王新恒[1] 王倩云 王佳杰 赵国锋[1] 靳文强 WANG Xin-heng;WANG Qian-yun;WANG Jia-jie;ZHAO Guo-feng;JIN Wen-qiang(School of Communication, Chongqing University of Posts and Telecommunications, Chongqing 400065, China)

机构地区：[1]重庆邮电大学通信学院,重庆400065

出　　处：《通信学报》2017年第9期95-105,共11页Journal on Communications

基　　金：国家自然科学基金资助项目(No.61402065);国家重点基础研究发展计划("973"计划)基金资助项目(No.2012CB315803;No.2012CB315806)~~

摘　　要：为了对软件定义无线网络系统中虚拟接入点(VAP)状态信息进行实时测量,根据实际网络中虚拟接入点性能的数据特征,提出一种基于随机游走的自适应矩阵填充算法(RW-MC)。首先,基于离散度和覆盖度的采样模型确定初始样本点;然后,利用随机游走模型对之前时隙的采样点序列建模分析,确定新时隙的测量点;最后,比较相邻窗口的恢复矩阵中重叠部分的误差率与标准误差,实现测量点的动态自适应。实验表明,该测量方法能够在低采样率、低误差的情况下实现对全网VAP实时感知。Concerning the continually perceiving performance of virtual access points(VAP)was urgent in software-defined wireless network(SDWN),with the features of VAPs'measurement data(VMD),a self-adaptive matrix completion algorithm based on random walk was proposed,named RW-MC.Firstly,the discrete ratio and covering ratio of VMD account for a sample determination model was used to claim initial samples.Secondly,random walk model was implemented for generating sampling data points in the next iteration.Finally,a self-adaptive sampling redress model concerning the differences between the current error rates and normalize error rates of neighboring completion matrices.The experiments show that the approach can collect the real-time sensory data,meanwhile,maintain a relatively low error rate for a small sampling rate.

关 键 词：SDWN 矩阵填充 RW-MC 随机游走 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]