[1]何 清,史忠植.基于超曲面的分类算法研究进展[J].智能系统学报,2007,2(06):1-7.
 HE Qing,SHI Zhong-zhi.Research advances in classification algorithm based on hy per-surface[J].CAAI Transactions on Intelligent Systems,2007,2(06):1-7.
点击复制

基于超曲面的分类算法研究进展(/HTML)
分享到:

《智能系统学报》[ISSN:1673-4785/CN:23-1538/TP]

卷:
第2卷
期数:
2007年06期
页码:
1-7
栏目:
出版日期:
2007-12-25

文章信息/Info

Title:
Research advances in classification algorithm based on hy per-surface
文章编号:
1673-4785(2007)06-0001-07
作者:
何 清12史忠植12
1.中国科学院计算技术研究所,北京100080;
2.中国科学院研究生院,北京100 039
Author(s):
HE Qing12 SHI Zhong-zhi12
1.Institute of Computing Technology, Chinese Academy of Sciences, Beijing 1 00080,China;
2.Graduate University, Chinese Academy of Sciences, Beiji ng 100039, China
关键词:
超曲面分类算法机器学习
Keywords:
hypersurface classification algorithmmachine le arning
分类号:
TP301
文献标志码:
A
摘要:
综述了基于超曲面的分类算法,该算法通过区域合并计算获得多个超平面组成的双侧闭曲面作为分类超曲面对空间进行划分. 分类超曲面可以有效地解决在有限连通区域分布很复杂的非线性数据多类分类问题,分析了算法准确率与极小样本集的关系, 总结了已有成就和最新进展, 指出了基于超曲面的分类算法进一步发展的方向.
Abstract:
In this paper, a classification method based on Hyper Surface (“HSC” for short ) is introduced. In this method, the space is partitioned through classification h ypersurface which are doublesided closed surfaces consisting of several hype r surfaces by merging the connected regions. HSC can efficiently solve the nonline ar multiclass classification problems, in which the sample data distributions a re very complicated within the finite connected regions. The relationship betwee n accuracy and the minimal consistent subset is analyzed. Finally, the existing achievements and the latest progresses in this subject are summarized, and the future research directions are pointed out.

参考文献/References:

[1] VAPNIK V. The nature of statistical learning theory[M]. New York: Sprin gerVerlag, 1995. 
[2]VALIANT L G. A theory of the learnable[J]. Communications of the ACM, 1 984, 27(11): 1134-1142. 
[3]BLUMER A, EHRENFEUCHT A, HASSLER D, et al. Classifying learnable geometri c concepts with the VapnikChervonenkis dimension[A]. Proceedings of the 19th A nnual ACM Symposium on Theory of Computing[C]. Berkeley, US, 1986.
[4]DANA A . Computational learning theory: survey and selected bibliog ra phy[A]. Proceedings of the Twentyfourth Annual ACM Symposium on Theory of Co mputing[C]. [S.l.],1992.
[5]DUDA R O, HART P E. Pattern classification and scene analysis[M]. New Y ork: John Wiley & Sons, 1973. 
[6]DUDA R O, HART P E, STOCK D G. 模式分类[M]. 北京:机械工业出版社,2003. 
[7]MENDELSON S. A few notes on statistical learning theory[R]. Advanced Le ctures on Machine Learning, LNAI 2600, 2003.
[8】张文生,丁 辉,王 珏. 基于邻域原理计算海量数据支持向量的研究[J]. 软件学报,2001,12(5): 711-720. 
 ZHANG Wensheng, DING Hui, WANG Jue. Study on computing the support vectors of ma ssive data based on neighborhood principle[J]. Journal of Software, 2001, 12( 5):711-720.
[9]TAO Q, WANG J. Kernel projection algorithm for largescale SVM pro blems[J]. Journal of Computer Science Technology, 2002, 17 (5): 556-564.
[10] TAO Qing, WU Gaowei, WANG Jue. A generalized S–K algorithm for learning SVM classifiers[J]. Pattern Recognition Letters, 2004, 25(10): 1165-1171.
[11]许建华, 张学工, 李衍达. 一种基于核函数的非线性感知器算法[J]. 计算机学报, 2002,25(7): 689-695
. XU Jianhua, ZHANG Xuegong, LI Yanda. A nonlinear perceptron algorithm based on k ernel functions[J]. Chinese Journal of Computers, 2002, 25(7): 689-695.
[12]ZHANG Ling, ZHANG Bo. A geometrical representation of McCulloch P itts neu ral model and its applications[J]. IEEE Transactions on Neural Networks, 1999, 10(4): 925-929.
[13]吴 涛,张 铃,张燕平.机器学习中的核覆盖算法[J].计算机学报,2005,2 8(8):1295-1301.
 WU Tao, ZHANG Ling, ZHANG Yanping. Kernel covering algorithm for machine learnin g[J]. Chinese Journal of Computers, 2005, 28(8): 1295-1301.
[14]陶 品,张 钹,叶 榛.构造型神经网络双交叉覆盖增量学习算法[J]. 软件学报,2003,14(2):194-201.
 TAO Pin, ZHANG Bo, YE Zhen. An incremental bicovering learning algorithm for con structive neural network[J]. Journal of Software, 2003, 14(2): 194-201.
[15]张 铃,张 钹,殷海风. 多层前向网络的交叉覆盖设计算法[J]. 软件学报,1999,10(7):737-742. 
 ZHANG Ling, ZHANG Bo, YIN Haifeng. An alternative covering design algorithm of m ultilayer neural networks[J]. Journal of Software, 1999, 10(7): 737-742.
[16] 周志华,曹存根.神经网络及其应用[M].北京:清华大学出版社,2004.
[17]王守觉.仿生模式识别(拓扑模式识别) ——一种模式识别新模型的理论与应用[J]. 电子学报, 2002,30(10): 1417-1420. 
WANG Shoujue. Bionic (topological) pattern recognition—a new model of patter n recognition theory and its applications[J]. Acta Electronica Sinica, 2002, 30(10): 1417-1420.
[18]WANG S J, QU Y F, LI W J. Face recognition: biomimetic pattern re cognition vs traditional recognition[J]. Acta Electronica Sinica, 2004, 32(7): 1057-10 61. 
[19]CAO W M, HAO F, WANG S J. The application of DBF neural networks for obj ec t recognition[J]. Information SciencesInformatics and Computer Science: An I nternational Journal, 2004, 160(1-4): 153-160
[20]王守觉,王柏南.人工神经网络的多维空间几何分析及其理论[J].电子学报 ,2002,30(1):1-4. 
 WANG Shoujue, WANG Bainan. Analysis and theory of highdimension space geometry for artificial neural networks[J]. Acta Electronica Sinica,2002, 30(1): 1-4 
[21]XU Zongben, MENG Deyu, JING Wenfeng. A new approach for classific ation: vi sual simulation point of view[A]. Proceedings of the ISNN 2005[C]. Spring Verlag, 2005.
[22]张 涌,朱 洪.一类弱集合覆盖问题的近似算法[J]. 计算机学报, 2005, 28(9):1497-1500.
ZHANG Yong, ZHU Hong. Approximation algorithms for the problems of weak set cove r[J]. Chinese Journal of Computers, 2005, 28(9): 1497-1500
[23]HE Qing, SHI Zhongzhi, REN Li’an. The classification method base d on hype r surface[A].Proceedings of the IEEE International Joint Conference on Neural Networks[C].Hawaii, 2002.
[24]HE Qing, SHI Zhongzhi, REN Li’an. The multiclass classification metho d i n large database based on hyper surface[A]. Proceedings of the International Conference on Machine Learning and Application[C]. Las Vegas, 2002.
[25]何 清,任力安,史忠植. 基于超曲面的海量数据直接分类法[J].计算机学报,2003,26(2):206-211. 
 HE Qing, REN Li’an, SHI Zhongzhi. The large data direct classifying method base d on hyper surface[J]. Chinese Journal of Computers, 2003, 26(2):206-211.
[26]HE Qing, Shi Zhongzhi, REN Li’an,et al. A novel classification m ethod bas ed on hypersurface[J]. International Journal of Mathematical and Computer Mode ling, 2003, 38:395-407.
[27]HE Qing, ZHAO Xiurong, SHI Zhongzhi. Clssification based on dimension tr an sposition for high dimention data[J]. Soft Computing, 2007, 11(4): 329-334.
[28]ZHAO Xiurong, HE Qing, SHI Zhongzhi. Hyper surface classifier ensemble fo r high dimensional data sets[A]. Proceedings of the 3rd International Symposium on Neural Networks (ISNN 2006)[C]. Berlin: SpringerVerlag, 2006.
[29]DASARATHY B V. Minimal consistent set (MCS) identification for optimal n ea rest neighbor decision systems design [J]. IEEE Trans Syst, Man, Cybern, 199 4,2 4 (3) :511 - 517.
[30]HANSEN L K, SALAMON P.Neural network ensembles[J]. IEEE Trans Pattern Analysis and Machine Intelligence, 1990,12(10): 993-1001.
[31]ZHOU Zhihua WU Jianxin JIANG Yuan,et al.Genetic algorithm based selectiv e neural network ensemble[A]. Proceedings of the 17th International Joint Confe rence on Artificial Intelligence[C].Seattle, US, 2001.
[32]ZHOU Z H, WU J, TANG W. Ensembling neural networks: many could be better than all [J]. Artificial Intelligence, 2002, 137(1-2): 239-263.
[33]GALLANT S I. Connectionist expert systems[J]. Communications of the AC M, 1988, 31(2): 152-169. 
[34]SETIONO R, LIU H. Understanding neural networks via rule extraction[A]. Pr oceedings of the 14th International Joint Conference on Artificial Intelligence [C]. Montreal, Canada, 1995. 
[35]WU X. Knowledge acquisition from databases[M]. Norwood, 1995. 
[36]ALEXANDER J A, MOZER M C. Templatebased procedures for neural network interpretation[J]. Neural Networks, 1999, 12(3): 479-498
[37]周志华,曹存根.神经网络及其应用[M].北京:清华大学出版社,2004.
[38】ZHOU Z H. Rule extraction: using neural networks or for neural ne tworks[J].Journal of Computer Science and Technology, 2004, 19(2): 24-253.
[39]ZHOU Z H, JIANG Y, CHEN S F. A general neural framework for classificati on rule mining[J]. International Journal of Computers, Systems and Signals, 2000 , 1(2): 154-168
[40】周志华, 陈世福. 神经网络规则抽取[J]. 计算机研究与发展, 2002, 39(4 ): 398-405.
 ZHOU Zhihua, CHEN Shifu. Rule extraction from neutral networks[J]. Journal of Computers Research and Development, 2002, 39(4):398-405.
[41]周志华, 陈世福. 神经网络集成[J]. 计算机学报, 2002, 25(1): 1-8.
ZHOU Zhihua, CHEN Shifu. Neural network ensemble[J]. Chinese Jornal of Comput ers,2002, 25(1):1-8.
[42]ZHOU Z H, JIANG Y, CHEN S F. Extracting symbolic rules from train ed neural networks[J]. AI Communications, 2003, 16(1): 3-15.
[43]ANDREWS R, DIEDERICH J, TICKLE A B. Survey and critique of techniques fo r extracting rules from trained artificial neural networks[J]. KnowledgeBased Systems, 1995, 8(6): 373-389.
[44]JOHN S T,ROBERT C,WILLIAMSON P. Generalization performance of classifier s in terms of observed covering numbers fischer[A]. Proceedings of the EuroC OLT’99[C]. Berlin: SpringerVerlag, 1999.
[45]王 珏,石纯一. 机器学习研究[J].广西师范大学学报(自然科学版), 2003 ,21(2):1-15. 
 WANG Jue, SHI Chunyi. Investigations on machine learning[J]. Journal of Guangx i Nor mal University (Natural Science Edition), 2003, 21(2):1-15.
[46]HE Qing, ZHAO Xiurong, SHI Zhongzhi. Sampling based on minimal consisten t subset for hyper surface classfication [A]. Proceedings of the Sixth Interna tional Cenference on Machine Learning and Cybernetics [C]. Hong Kong, 2007.
[47]HE Qing, ZHUANG Fuzhen, ZHAO Xiurong, et al. Enhanced algorit h m performance for classification based on hyper surface using bagging and adaboost [A]. Proceedings of the Sixth International Conference on Machine Learning and C ybernetics[C]. Hong Kong, 2007.
[48]HE Qing, ZHAO Xiurong, SHI Zhongzhi . A cognitive data visualization m et hod based on hyped surface[A]. Proc 6th IEEE Int Conf on Cognitive Informatics (ICCIO07)[C]. [S.l.], 2007.
[49]HE Qing, ZHAO Xiurong, LUO Ping,et al. Combination methodologies of m ul tiAgent hyper surface classifiers: design and implementation issues [A]. LNA I 4476[C].Berlin: SpringerVerlag, 2007.

相似文献/References:

[1]赵春晖,张燚.相关向量机分类方法的研究进展与分析[J].智能系统学报,2012,7(04):294.
 ZHAO Chunhui,ZHANG Yi.Research progress and analysis on methods for classification of RVM[J].CAAI Transactions on Intelligent Systems,2012,7(06):294.
[2]郭龙伟,关欣,李锵.基于测度学习支持向量机的钢琴乐谱难度等级识别[J].智能系统学报,2018,13(02):196.[doi:10.11992/tis.201612012]
 GUO Longwei,GUAN Xin,LI Qiang.Recognition of difficulty level of piano score based on metric learning support vector machine[J].CAAI Transactions on Intelligent Systems,2018,13(06):196.[doi:10.11992/tis.201612012]
[3]程麟焰,胡峰.基于模糊超网络的知识获取方法研究[J].智能系统学报,2019,14(03):479.[doi:10.11992/tis.201804055]
 CHENG Linyan,HU Feng.Fuzzy hypernetwork-based knowledge acquisition method[J].CAAI Transactions on Intelligent Systems,2019,14(06):479.[doi:10.11992/tis.201804055]
[4]刘牧雷,徐菲菲.基于三支决策的序列数据代价敏感分类算法[J].智能系统学报,2019,14(06):1255.[doi:10.11992/tis.201905049]
 LIU Mulei,XU Feifei.A sequence data, cost-sensitive classification algorithm based on three-way decisions[J].CAAI Transactions on Intelligent Systems,2019,14(06):1255.[doi:10.11992/tis.201905049]

备注/Memo

备注/Memo:
收稿日期:2007-01-10.
基金项目:
国家自然科学基金资助项目(60435010,60675010);
国家重点基础研究发展计划资助项目(2006AA01Z128);
北京市自然科学基金资助项目(40520 25).
作者简介:
 何 清, 1965年生,教授,博士生导师,主要研究方向为人工智能、数据挖掘、机器学习、模糊集理论,发表学术论文60多篇.
E-mail:heq@ics.ict.ac.cn.
史忠植,1944年生,研究员,博士生导师,主要研究方向为人工智能、多主体系统、数据挖掘、机器学习、知识工程等.1979年、1998年、2001年均获中国科学院科技进步二等奖,199 4年获中国科学院科技进步特等奖,2002年获国家科技进步二等奖,发表学术论文400多篇,出版专著5部.
 E-mail:shizz@ics.ict.ac.cn.
更新日期/Last Update: 2009-05-08