您的位置:山东大学 -> 科技期刊社 -> 《山东大学学报(理学版)》

山东大学学报(理学版) ›› 2016, Vol. 51 ›› Issue (9): 137-144.doi: 10.6040/j.issn.1671-9352.3.2015.064

• • 上一篇    下一篇

基于筛选机制的L1核学习机分布式训练方法

及歆荣1,2,侯翠琴1,侯义斌1,赵斌3   

  1. 1.北京市物联网软件与系统工程技术研究中心, 北京 100124;2.河北工程大学信息与电气工程学院, 河北 邯郸 056038;3.北京工业大学软件学院, 北京 100124
  • 收稿日期:2015-08-02 出版日期:2016-09-20 发布日期:2016-09-23
  • 作者简介:及歆荣(1978— ),女,博士研究生,讲师,研究方向为分布式机器学习、无线传感器网络. E-mail:2596008@qq.com
  • 基金资助:
    国家自然科学基金青年基金资助项目(61203377)

A distributed training method for L1 regularized kernel machines based on filtering mechanism

JI Xin-rong1,2, HOU Cui-qin1, HOU Yi-bin1, ZHAO Bin3   

  1. 1. Beijing Engineering Research Center for IoT Software and Systems, Beijing 100124, China;
    2. School of Information and Electrical Engineering, Hebei University of Engineering, Handan 056038, Hebei, China;
    3. School of Software Engineering, Beijing University of Technology, Beijing 100124, China
  • Received:2015-08-02 Online:2016-09-20 Published:2016-09-23

摘要: 为降低无线传感器网络中核学习机训练时的数据通信代价和节点计算代价,研究了基于筛选机制的L1正则化核学习机分布式训练方法。提出了一种节点局部训练样本筛选机制,各节点利用筛选出的训练样本,在节点模型对本地训练样本的预测值与邻居节点间局部最优模型对本地训练样本预测值相一致的约束下,利用增广拉格朗日乘子法求解L1正则化核学习机分布式优化问题,利用交替方向乘子法求解节点本地的L1正则化核学习机的稀疏模型;仅依靠相邻节点间传输稀疏模型的协作方式,进一步优化节点局部模型,直至各节点模型收敛。基于此方法,提出了基于筛选机制的L1正则化核最小平方误差学习机的分布式训练算法。仿真实验验证了该算法在模型预测正确率、模型稀疏率、数据传输量和参与模型训练样本量上的有效性和优势。

关键词: 无线传感器网络, 分布式学习, 样本筛选机制, 增广拉格朗日乘子法, L1正则化, 核学习机

Abstract: To decrease the amount of data transferred and the computing cost during training a kernel machine in wireless sensor network, a distributed training method for L1-regularized Kernel Minimum Square Error machine based on filtering mechanism was proposed. First, filtering mechanism of samples was presented and used on each node. Second, with consistency constraint on the local model of each node and its local optimal one obtained by exchanging the local model with its all neighbours, the distributed optimization problem of L1-regularized Kernel Minimum Square Error machine was solved by Augmented Lagrange Method of Multipliers, and the local optimization problem of L1-regularized Kernel Minimum Square Error machine on each node was solved by Alternating Direction Method of Multipliers. Then, the spares model obtained on each node was transferred to its all neighbor nodes. This process iterates until the local model on each node converges. For carrying out this method,a novel distributed training algorithm for L1-regularized Kernel Minimum Square Error based on filtering of samples was proposed. Simulation results prove the validity of the proposed algorithm in terms of convergence, sparse rate of model, the amount of data transferred and the number of samples used in model training.

Key words: wireless sensor network, L1-regularized, augmented Lagrange method of multipliers, kernel machines, distributed learning, filtering mechanism of samples

中图分类号: 

  • TP181
[1] 吕方旭,张金成,郭相科,等. 基于WSN的战场声目标多特征联合智能分类识别[J]. 科学技术与工程, 2013, 13(35):10713-10721. LÜ Fangxu, ZHANG Jincheng, GUO Xiangke, et al. Multi feature joint intelligent classification and recognition of battlefield acoustic targets based on WSN[J]. Science Technology and Engineering, 2013, 13(35):10713-10721.
[2] TAGHVAEEYAN S, RAJAMANI R. Portable roadside sensors for vehicle counting, classification, and speed measurement[J]. IEEE Transactions on Intelligent Transportation Systems, 2014, 15(1):73-83.
[3] SHAHID N, NAQVI I H, QAISAR BIN S. Quarter-sphere svm: attribute and spatio-temporal correlations based outlier and event detection in wireless sensor networks[J]. IEEE Wireless Communication and Networking Conference: Mobile and Wireless Networks, 2012: 2048-2053.
[4] 刘倩,崔晨,周杭霞. 改进型SVM多类分类算法在无线传感器网络中的应用[J].中国计量学院学报, 2013, 24(3):298-303. LIU Qian, CUI Chen, ZHOU Hangxia. Application of improved SVM multi class classification algorithm in wireless sensor networks[J]. Journal of China Jiliang University, 2013, 24(3):298-303.
[5] 韩屏,李方敏,罗婷. 无线传感器网络的分布式目标跟踪算法[J]. 北京邮电大学学报, 2009, 32(1):90-94. HAN Ping, LI Fangmin, LUO Ting. Distributed target tracking algorithm for wireless sensor networks[J]. Journal of Beijing University of Posts and Telecommunications, 2009, 32(1):90-94.
[6] RAJ A B, RAMESH M V, KULKARNI R V, et al. Security enhancement in wireless sensor networks using machine learning[C] // Proceedings of IEEE 14th International Conference on High Performance Computing and Communications. New York: ACM, 2012: 1264-1269.
[7] SCHÖLKOPF B, SMOLA A. Learning with kernels: support vector machines, regularization, optimization and beyond[J]. MIT Press, 2002: 61-118.
[8] 及歆荣,侯翠琴,侯义斌. 无线传感器网络下线性支持向量机分布式协同训练方法研究[J].电子与信息学报,2015, 37(3):708-714. JI Xinrong, HOU Cuiqin, HOU Yibin. Research on distributed cooperative training method for linear support vector machine in wireless sensor networks[J]. Journal of Electronics and Information, 2015, 37(3):708-714.
[9] GUESTRIN C, BODIK P, THIBAUX R, et al. Distributed regression: an efficient framework for modeling sensor network data[C] // Proceedings of the International Conference on Information Processing in Sensor Networks. New York: ACM, 2004: 1-10.
[10] PREDD J B, KULKARNI S R, POOR H V. Distributed kernel regression: an algorithm for training collaboratively[C] // Proceedings of IEEE Information Theory Workshop. New York: IEEE, 2006: 332-336.
[11] PREDD J B, KULKARNI S R, POOR H V. A Collaborative training algorithm for distributed learning[J]. IEEE Transactions on Information Theory, 2009, 55(4):1856-1870.
[12] FLOURI K, BEFERULL-LOZANO B, TSAKALIDES P. Optimal gossip algorithm for distributed consensus SVM training in wireless sensor networks[C] // Proceedings of the 16th International Conference on Digital Signal Processin(DSP '09). Piscataway: IEEE, 2009: 886-891.
[13] FLOURI K, BEFERULL-LOZANO B, TSAKALIDES P.Training a support-vector machine-based classifier in distributed sensor networks[C] // Proceedings of the 14nd European Signal Processing Conference. Florence:[S.l.] , 2006: 4-8.
[14] YUMAO L, ROYCHOWDHURY V, VANDENBERGHE L. Distributed parallel support vector machines in strongly connected networks[J]. IEEE Transactions on Neural Networks, 2008, 19(7):1167-1178.
[15] XU Jianhua, ZHANG Xuegong, LI Yanda. Kernel MSE algorithm: a unified framework for KFD, LS-SVM and KRR[C] // Proceedings of International Joint Conference on Neural Networks. Piscataway: IEEE, 2001: 1486-1491.
[16] BOYD S, PARIKH N, CHU E. Distributed optimization and statistical learning via the alternating direction method of multipliers[J]. Foundations and Trends in Machine Learning, 2011, 3(1):1-122.
[17] BERTSEKAS D P, TSITSIKLIS J N. Parallel and distributed computation: numerical methods[M]. London: Prentice-Hall, 1997: 243-247.
[1] 周先存, 黎明曦, 李瑞霞, 徐明鹃, 凌海波. 多点协作复制攻击检测研究[J]. 山东大学学报(理学版), 2015, 50(07): 54-65.
[2] 万智萍. 一种混合Das协议的无线传感器网络身份验证协议[J]. 山东大学学报(理学版), 2015, 50(05): 12-17.
[3] 张晶, 薛冷, 崔毅, 容会, 王剑平. 基于无线传感器网络的双混沌数据加密算法建模与评价[J]. 山东大学学报(理学版), 2015, 50(03): 1-5.
[4] 李国庆. 无线传感器网络基于联系数的信任评估模型[J]. 山东大学学报(理学版), 2014, 49(09): 123-128.
[5] 赵泽茂1,刘洋1,张帆1,2,周建钦1,张品1. 基于角度和概率的WSN源位置隐私保护路由研究[J]. J4, 2013, 48(09): 1-9.
[6] 万润泽1,雷建军1,袁操2. 基于模糊聚类理论的无线传感器节点休眠优化策略[J]. J4, 2013, 48(09): 17-21.
[7] 吕家亮1,2,3,王英龙1,3,崔焕庆1,魏诺2,3,郭强2,3. 基于微粒群优化的三维无线传感网定位算法研究[J]. J4, 2013, 48(05): 78-82.
[8] 崔焕庆1,2,王英龙1*,吕家亮1,2,魏诺1. 基于随机微粒群算法的分布式节点定位方法[J]. J4, 2012, 47(9): 51-55.
[9] 郭晓东1,杜鹏1,张雪芬2. 一种WSN中能量有效的分布式检测和功率分配算法[J]. J4, 2012, 47(9): 60-64.
[10] 刘梦君,刘树波*,刘泓晖,蔡朝晖,涂国庆. 异构无线传感器网络动态混合密钥管理方案研究[J]. J4, 2012, 47(11): 67-73.
[11] 姚武军1,丁谊1,魏立线1,杨晓元1,2. 基于RSSI的无线传感器网络ARQ差错控制策略[J]. J4, 2011, 46(9): 61-66.
[12] 李桂青,高仲合,王楠楠. 基于簇头集的无线传感器网络定向扩散协议[J]. J4, 2010, 45(11): 37-42.
[13] 周书旺1,2,3,王英龙1,3,郭强1,2,魏诺1,2,郭文娟1,3. 一种考虑可能区域和智能搜索相结合的定位算法[J]. J4, 2010, 45(11): 27-31.
[14] 郭文娟1,2,王英龙1,2,魏诺1,3,郭强1,3,周书旺1,2,3. 基于卡尔曼滤波的无线传感器网络时钟同步协议[J]. J4, 2010, 45(11): 32-36.
[15] 周书旺 王英龙 郭强 魏诺. 基于微粒群算法的无线传感器网络节点定位方法[J]. J4, 2009, 44(9): 52-55.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!