基于SRM自组织多区域覆盖的可拒绝近邻分类算法研究

胡正平; 贾千文

doi:10.3724/SP.J.1146.2007.01360

基于SRM自组织多区域覆盖的可拒绝近邻分类算法研究

doi: 10.3724/SP.J.1146.2007.01360

基金项目:

河北省自然科学基金(F2008000891)，燕山大学博士基金项目(B287)和北京大学视觉与听觉信息处理国家重点实验室开放基金 (0507)资助课题

计量
- 文章访问数: 3460
- HTML全文浏览量: 104
- PDF下载量: 924
- 被引次数: 0
出版历程
- 收稿日期: 2007-08-27
- 修回日期: 2008-05-11
- 刊出日期: 2009-02-19

Rejecting Nearest Neighbor Classifier Based on Structural Risk Minimization Principle Self-organization Multiple Region Covering Model

摘要

摘要: 该文依据区分与划分相结合的可拒绝模式识别思路，提出了高维空间海量训练样本情况下的基于结构风险最小化决策的自组织多区域多球覆盖可拒绝近邻分类算法。该方法利用同类样本之间相互接近的特性，通过结构风险最小化紧几何覆盖策略，选择训练样本，通过自组织多区域多球覆盖模型构成同类样本的划分性描述，达到拒绝识别非训练类样本的目的，最后通过k近邻相互区分性比较确定真实类别。仿真实验结果表明该文的思路是合理可行的，在实际应用领域具有一定价值。
- 可拒绝模式识别;结构风险最小化原理;近邻分类
Abstract: According to the mode of rejecting pattern recognition principles, which is based on matter description and matter separation in uniform statistical pattern recognition, a rejecting nearest neighbor classifier based on structural risk minimization self-organization multiple region covering model is presented in this paper. This new model is better closer to the actual instance, rather than traditional statistical pattern recognition only using optimal separationas its main principle. Firstly, the optimal samples are selected from the training samples based on structural risk minimization, which is used for same class pattern matter description. Then the kNN distinguish is as a following step to identified the exact class. The simulation experimental results show that this method is valid and efficient.

HTML全文

参考文献(1)

王守觉. 仿生模式识别一种模式识别的新模型的理论与方法[J]. 电子学报, 2002, 30(10): 1418-1421.Wang Shou-jue. Bionic (topological) pattern recognitionA new model of pattern recognition theory and itsapplications. Acta Electronica Sinica, 2002, 30(10):1417-1420.[2]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, 28(8): 1295-1301.Wu Tao, Zhang Ling, and Zhang Yan-ping. Kernel coveringalgorithm for machine learning. Chinese Journal ofComputers, 2005, 28(8): 1295-1301.[3]赵莹, 高隽, 汪荣贵, 胡静. 一种新的广义近邻方法研究[J].电子学报, 2004, 32(F12): 196-198.Zhao Ying, Gao Jun, Wang Rong-gui, and Hu Jing. Extendednearest neighbor method based on bionic pattern recognition.Acta Electronica Sinica, 2004, 32(F12): 196-198.[4]David M J Tax and Robert P W Duin. Support vector datadescription [J].Machine Learning.2004, 54(1):45-66[5]Lee Ki Young, Kim Dae-Won, Lee Kwang H, and LeeDoheon. Density-induced support vector data description[J].IEEE Trans. on Neural Networks.2007, 18(1):284-289[6]Yen Chen-wen, Young Chieh-Neng, and Nagurka Mark L. Afalse acceptance error controlling method for hypersphericalclassifiers[J].Neurocomputing.2004, 57(1-4):295-312[7]Amit B, Philippe B, and Chris D. A support vector methodfor anomaly detection in hyperspectral imagery[J].IEEE Trans.on Geoscience and Remote Sensing.2006, 44(8):2282-2291[8]Rizvi S A, Saasawi T N, and Nasrabadi N M. A clutterrejection technique for FLIR imagery using region basedprincipal component analysis [J].Pattern Recognition.2000,33(11):1931-1933[9]Yang G Z and Huang T S. Human face detection in a complexbackground [J]. Pattern Recognition, 1994, 27(1): 58-63.[10]Cabrera J B D. On the impact of fusion strategies onclassification errors for large ensembles of classifiers. PatternRecognition, 2006, 39(11): 1963-1978.[11]娄震, 金忠, 杨静宇. 基于类条件置信变换的后验概率估计方法[J]. 计算机学报, 2005, 28(1): 18-24.Lou Zhen, Jin Zhong, and Yang Jing-yu. Novel approach toestimate posterior probabilities by class-conditionalconfidence transformations. Chinese Journal of Computers,2005, 28(1): 18-24.

施引文献

资源附件(0)

访问统计