基于AUC优化的非线性主动学习算法及其在障碍物检测中的应用

摘要
图/表
参考文献
相关文章 (9)

全文: PDF (938 KB) HTML ( ) 输出: BibTeX \| EndNote (RIS)
引用本文:
韩光, 赵春霞, 胡雪蕾. 基于AUC优化的非线性主动学习算法及其在障碍物检测中的应用[J]. 机器人, 2010, 32(3): 344-351.. HAN Guang, ZHAO Chunxia, HU Xuelei. A Nonlinear Active Learning Based on AUC Optimization and Its Application to Obstacle Detection. ROBOT, 2010, 32(3): 344-351..

摘要针对障碍物检测中因样本量过大而造成的标记困难以及样本分布不均衡等问题，提出了一种基于AUC优化的非线性主动学习算法．该算法的计算处理过程是：首先利用基于AUC优化的算法在训练集上对非线性分类器进行训练；然后利用已训练好的分类器对所有未标记样本进行分类；接着利用基于AUC优化的样本选择函数计算分类后的样本的得分；最后算法根据分值大小选出最有信息量样本，并且专家根据该样本所在的图像及在图像中位置对其进行标记并放入训练集中．重复上述过程，直到AUC收敛为止．在户外环境图像库上进行了实验，结果表明：该算法能显著减小数据标记的工作量，并能解决因样本分布不平衡而引起的次优解问题，与已有主动学习算法相比性能更优．

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	韩光
	赵春霞
	胡雪蕾

关键词 ：障碍物检测, 主动学习, AUC优化, 非线性分类器, 梯度下降法

Abstract：Aiming at difficulties in labeling caused by a large number of the samples,as well as uneven distribution of the samples in obstacle detection,a nonlinear active learning algorithm based on AUC(area under the receiver operating characteristic) optimization is proposed.Calculation process of this algorithm is as following.Firstly the AUC optimization method is used to train the nonlinear classifier on the training set.Then all the unlabeled samples are classified with the trained classifier.Secondly all the classified samples are scored using the sample selection function based on AUC optimization,and then the best representative samples are selected according to the scores.Finally these samples are labeled by the expert based on the images and location in the image,and then all the labeled samples are put in the training set.The above process is repeated until the AUC converges.Experiments are performed in outdoor environment image database.Experimental results demonstrate that the proposed algorithm can significantly reduce the workload of labeling the samples,and can solve the problem of the sub-optimal solution caused by the uneven sample distribution.The performance is also better than the other active learning algorithms.

Key words： obstacle detection active learning AUC optimization nonlinear classifier gradient descent method

收稿日期: 2009-06-11 录用日期: 2010-02-25

:	TP24
	TP18

基金资助:国家自然科学基金资助项目（60705020,90820306）

通讯作者: 韩光 E-mail: hanguang8848@163.com

作者简介: 韩光(1981- )，男，博士生．研究领域：模式识别，智能机器人，机器学习．
赵春霞(1964- )，女，教授，博士生导师．研究领域：模式识别，图像处理，人工智能和智能机器人．
胡雪蕾(1977- )，女，副教授，硕士生导师．研究领域：机器学习，神经网络，模式识别和智能机器人．

链接本文:

https://robot.sia.cn/CN/ 或 https://robot.sia.cn/CN/Y2010/V32/I3/344

[1] Thrun S,Montemerlo M,Dahlkamp H,et al.Stanley:The robot that won the DARPA grand challenge[J].Journal of Field Robotics,2006,23(9):661-692.

[2] Dima C,Hebert M.Active learning for outdoor obstacle de tection[C]//Proceedings of the Robotics:Science and Systems.Cambridge,Massachusetts,USA:MIT Press,2005:9-16.
[3] Manduchi R,Castano A,Talukder A,et al.Obstacle detection and terrain classification for autonomous off-road navigation[J].Autonomous Robots,2005,18(1):81-102.

[4] Dima C,Hebert M,Stentz A.Enabling learning from large datasets:Applying active learning to mobile robotics[C]//IEEE International Conference on Robotics and Automation.Piscat away,NJ,USA:IEEE,2004:108-114.
[5] van der Mark W,van den Heuvel J C,Groen F C A.Stereo based obstacle detection with uncertainty in rough terrain[C]//Proceedings of the IEEE Intelligent Vehicles Symposium.Pis caraway,NJ,USA:IEEE,2007:1005-1012.
[6] 张健沛,徐华.支持向量机(SVM)主动学习方法研究与应用[J].计算机应用,2004,24(1):1-3.Zhang Jianpei,Xu Hun.Study and application of active learning with SVM[J].Journal of Computer Applications,2004,24(1):1-3.
[7] Tong S,Koller D.Support vector machine active learning with applications to text classification[J].Journal of Machine Learn ing Research,2002,2(1):45-66.
[8] Seang H S,Opper M,Sompolinsky H.Query by commit tee[C]//Annual ACM Workshop on Computational Learning Theory.New York,USA:ACM,1992:287-294.
[9] Breiman L.Bagging predictors[J].Machine Learning,1996,24(2):123-140.
[10] Abe N,Mamitsuka H.Query learning strategies using boost ing and bagging[C]//15th International Conference on Machine Learning.San Fransisco,USA:Morgan Kanfmann,1998:1-9.
[11] Lewis D D,Gale W A.A sequential algorithm for training text classifiers[C]//Annual ACM SIGIR Conference on Research and Development in Information Retrieval.New York,USA:Springer-Verlag,1994:3-12.
[12] Vlachos A.A stopping criterion for active learning[J].Com puter Speech and Language,2008,22(3):295-312.

[13] Cord M,Gosselin P H,Philipp-Foliguet S.Stochastic explo ration and active learning for image retrieval[J].Image and Vi sion Computing,2007,25(1):14-23.
[14] 田春娜,高新波,李沽.基于嵌入式Bootstrap的主动学习示例选择方法[J].计算机研究与发展,2006,43(10):1706-1712.Tian Chunna,Gao Xinbo,Li Jie.An example selection method for active learning based on embedded bootstrap algorithm[J].Journal of Computer Research and Development,2006,43(10):1706-1712.
[15] Yan R,Yang J,Hauptmann A.Automatically labeling video data using multi-class active learning[C]//IEEE International Conference on Computer Vision.Piscataway,NJ,USA:IEEE,2003:516-523.
[16] Doamez,P,Carbonell,J G.Active sampling for rank learn ing via optimizing the area under the ROC curve[C]//31st Eu ropean Conference on Information Retrieval Research.Berlin,Germany:Springer-Verlag,2009:78-89.
[17] Marrocco C,Duin R P W,Tortorella F.Maximizing the area un der the ROC curve by pairwise feature combination[J].Pattern Recognition,2008,41(6):1961-1974.

[18] Cortes C,Mohri M.AUC optimization vs.error rate minimiza tion[M]//Advances in Neural Information Processing Systems,vol.16.Cambridge,MA,USA:MIT Press,2004:313-320.
[19] Huang J,Ling C X.Using AUC and accuracy in evaluating learning algorithms[J].IEEE Transactions on Knowledge and Data Engineering,2005,17(3):299-310.

[20] Calders T,Jaroszewicz S.Efficient AUC optimization for classi fication[C]//11th European Conference on Principles and Prac rice of Knowledge Discovery in Databases.Berlin,Germany:Springer-Verlag,2007:42-53.
[21] Campbell C,Cristianini N,Smola A J.Query learning with large margin classifiers[C]//17th International Conference on Machine Learning.San Francisco,CA,USA:Morgan Kanf mann,2000:111-118.
[22] Pietik(a)inen M,Nurmela T,M(a)enp(aa) T,et al.View-based recog nition of real-world textures[J].Pattern Recognition,2004,37(2):313-323.

[23] Permuter H,Francos J,Jermyn J.A study of Ganssian mixture models of color and texture features for image classification and segmentation[J].Pattern Recognition,2006,39(4):695-706.

[24] Castano R,Manduchi R,Fox J.Classification experiments on real-world texture[C]//3rd Workshop on Empirical Evaluation Methods in Computer Vision.Pasadena,CA,USA:JPL,2001:3-20.
[25] Yan L,Dodier R,Mozer M C,et al.Optimizing classifier per formance via an approximation to the Wilcoxon-Mann-Whitney statistic[C]//20th International Conference on Machine Learn ing.Menlo Park,CA,USA:AAAI,2003:848-855.