覆盖算法的概率模型
作者:
基金项目:

Supported by the National Natural Science Foundation of China under Grant Nos.60475017, 60675031 (国家自然科学基金); the National Basic Research Program of China under Grant No.2004CB318108 (国家重点基础研究发展计划(973)); the National Research Foundation for the Doctoral Program of Higher Education of the Ministry of Education of China under Grant No.20040357002 (国家教育部高等学校博士学科点专项科研基金); the Natural Science Foundation of Anhui Province of China under Grant No.050420208 (安徽省自然科学基金); the Provincial Project of Natural Science Research for Colleges and Universities of Anhui Province of China under Grant Nos.2006KJ015A, 2006KJ244B (安徽高等学校省级自然科学研究项目); the Innovative Research Team of 211 Project in Anhui University of China (安徽大学学术创新团队); the Program of Superior Teacher Team in Anhui University of China (安徽大学人才队伍建设经费)

  • 摘要
  • | |
  • 访问统计
  • |
  • 参考文献 [25]
  • |
  • 相似文献 [20]
  • |
  • 引证文献
  • | |
  • 文章评论
    摘要:

    要从本质上提高覆盖算法的精度,必须在算法中引入全局的优化计算.为此,先将覆盖算法扩展成核覆盖算法(以高斯函数为核函数),再利用高斯函数的概率意义(高斯分布),为核覆盖算法建立一个有限混合概率模型,在此基础上,利用"最大似然原理"引入全局优化计算,并利用EM(expectation maximization)方法进行求解,完成对覆盖算法的全局优化计算,从而扩大覆盖方法的使用范围并提高算法的精度,且将它从确定的模型扩展成概率的模型,后者更具抗噪声干扰的能力.最后给出模拟实验,实验比较结果表明,经优化后的概率模型确实提高了算法的精度.

    Abstract:

    It is necessary to bring global optimization in covering algorithm to improve its precision of classification.So a probabilistic model of covering algorithm is put forward in this paper.Firstly,the covering algorithm is ameliorated to kernel covering model(Gaussian function is the kernel function),then a kind of finite mixture probabilistic model for kernel covering model is introduced according to the probabilistic meaning of Gaussian function.Finally,the global optimization calculation is inducted based on maximum likelihood theory and Expectation Maximization Algorithm.Therefore,the algorithm optimizes the covering network broadens the application domain of covering algorithm and improves its robustness.The experimental results show that the optimized probabilistic model of covering algorithm can improve the accuracy of classification.

    参考文献
    [1]Zhang L,Zhang B.A geometrical representation of McCulloch-Pitts neural model and its applications.IEEE Trans.on Neural Networks,1999,10(4):925-929.
    [2]Wu T,Zhang L,Zhang YP.Kernel covering algorithm for machine learning.Chinese Journal of Computers,2005,28(8):1295-1301 (in Chinese with English abstract).
    [3]Zhang YP,Zhang L,Wu T.A multi-side increase by degrees algorithm at machine learning.Acta Electronica Sinica,2005,33(2):327-331 (in Chinese with English abstract).
    [4]Ye SZ,Zhang B,Wu MR,Zhang WB.A fuzzy classifier based on the constructive covering approach in neural networks.Journal of Software,2003,14(3):429-434 (in Chinese with English abstract).http://www.jos.org.cn/1000-9825/14/429.htm
    [5]Tao P,Zhang B,Ye Z.An incremental BiCovering learning algorithm for constructive neural network.Journal of Software,2003,14(2):194-201 (in Chinese with English abstract).http://www.jos.org.cn/1000-9825/14/194.htm
    [6]Wu MR,Zhang B.A neural network algorithm for large scale pattern recognition problems.Journal of Software,2001,12(6):851-855 (in Chinese with English abstract).
    [7]Zhang M,Wu T,Wang LW,Cheng JX.The application of granularity of the quotient space theory in database and data warehouse.Journal of the Computer Engineering and Application,2003,39(17):47-49,60 (in Chinese with English abstract).
    [8]Wang LW,Wu T,Zhang M,Zhang L.Improvement on neighborhood covering algorithm and it's application.Pattern Recognition and Artificial Intelligence,2003,16(1):81-85 (in Chinese with English abstract).
    [9]Zhang YP,Zhang L,Wu T,Xu F,Zhang M,Wang LW.A structural learning algorithm based on covering algorithm and its application in stock forecasting.Journal of Computer Research and Development,2004,41(6):979-984 (in Chinese with English abstract).
    [10]Zhang YP,Zhang L,Duan Z.A constructive kernel covering algorithm and applying it to image recognition.Journal of Image and Graphics,2004,9(11):1304-1308 (in Chinese with English abstract).
    [11]Zhou Y.A study of algorithm of neural networks as classifiers and their application in text classification[Ph.D.Thesis].Hefei:Anhui University,2006 (in Chinese with English abstract).
    [12]Dempster AP,Laird NM,Rubin DB.Maximum likelihood from incomplete data using the EM algorithm (with discussion).Journal of the Royal Statistical Society (Series B),1977,39(1):1-38.
    [13]McLachlan GJ,Peel D.Finite Mixture Models.New York:John Wiley & Sons,2000.
    [14]Zhang L.Research on support vector machines and kernel methods[Ph.D.Thesis].Xi'an:Xi'an University of Electronic Sciences,2002 (in Chinese with English abstract).
    [2]吴涛,张铃,张燕平.机器学习中的核覆盖算法.计算机学报,2005,28(8):1295-1231.
    [3]张燕平,张铃,吴涛.机器学习中的多侧面递进算法MIDA.电子学报,2005,33(2):327-331.
    [4]叶少珍,张钹,吴鸣锐,郑文波.一种基于神经网络覆盖构造法的模糊分类器.软件学报,2003,14(3):429-434.http://www.jos.org.cn/ 1000-9825/14/429.htm
    [5]陶品,张钹,叶榛.构造型神经网络双交叉覆盖增量学习算法.软件学报,2003,14(2):194-201.http://www.jos.org.cn/1000-9825/14/ 194.htm
    [6]吴鸣锐,张钹.一种用于大规模模式识别问题的神经网络算法.软件学报,2001,12(6):851-855.
    [7]张旻,吴涛,王伦文,程家兴.商空间粒度计算理论在数据库和数据仓库中应用.计算机工程与应用,2003,39(17):47-49,60.
    [8]王伦文,吴涛,张旻,张铃.一种改进的领域覆盖算法与应用.模式识别与人工智能,2003,16(1):81-85.
    [9]张燕平,张铃,吴涛,徐锋,张旻,王伦文.基于覆盖的构造性学习算法(SLA)及在股票预测中的应用.计算机研究与发展,2004,41(6):979-984.
    [10]张燕平,张铃,段震.构造性核覆盖算法在图像识别中的应用.中国图像图形学报,2004,9(11):1304-1308.
    [11]周瑛.神经网络作为分类器的算法研究及在信息检索中的应用[博士学位论文].合肥:安徽大学,2006.
    [14]张莉.支撑矢量机与核方法研究[博士学位论文].西安:西安电子科技大学,2002.
    网友评论
    网友评论
    分享到微博
    发 布
引用本文

张铃,吴涛,周瑛,张燕平.覆盖算法的概率模型.软件学报,2007,18(11):2691-2699

复制
分享
文章指标
  • 点击次数:4832
  • 下载次数: 6486
  • HTML阅读次数: 0
  • 引用次数: 0
历史
  • 收稿日期:2006-04-24
  • 最后修改日期:2006-08-22
文章二维码
您是第19784169位访问者
版权所有:中国科学院软件研究所 京ICP备05046678号-3
地址:北京市海淀区中关村南四街4号,邮政编码:100190
电话:010-62562563 传真:010-62562533 Email:jos@iscas.ac.cn
技术支持:北京勤云科技发展有限公司

京公网安备 11040202500063号