期刊文献+

最大局部加权均值差异嵌入 被引量:4

Maximum Local Weighted Mean Discrepancy Embedding
在线阅读 下载PDF
导出
摘要 最大均值差异嵌入(Maximum Mean Discrepancy Embedding,MMDE)作为一种基于最大均值差异(MaximumMean Discrepancy,MMD)度量的特征提取方法被成功地运用.然而通过分析得知,该方法在处理原始输入空间上的特征提取问题时一定程度上缺乏适应性.因此本文在MMD准则的基础上,并结合已经被广泛研究和探讨的局部学习方法,提出一个新的评价度量:最大局部加权均值差异(Maximum LocalWeightedMean Discrepancy,MLMD),该度量反映源域和目标域分布差异时能充分考虑两个区域内在的局部结构,同时还能通过局部分布差异去反映全局分布差异.本文还在此度量的基础上提出一种能实现迁移学习任务并具有一定局部学习能力的特征提取方法:最大局部加权均值差异嵌入(Maximum Local WeightedMean Discrepancy Embedding,MWME).该方法不但能完成传统意义上的特征提取,同时还能完成在两个分布存在差异但相关的两个区域上实现领域适应学习,从而表明该特征提取方法具有较好的鲁棒性和适应性.实验证明MLMD准则和MWME方法具有上述优势. MMDE,regarded as a MMD-based feature extraction method,has been successfully used.However,when the feature extraction problems of the original input space have been solved,the MMDE lacks the suitability to some extent.Therefore,we propose Maximum Local Weighted Mean Discrepancy(MLMD)by integrating the theory and technique of local learning methods.The measurement considers fully the internal local structure between domains;at the same time,the global distribution discrepancy can be reflected by the local distribution discrepancy.We also,based on the above measurement,propose Maximum Local Weighted Mean Discrepancy Embedding(MWME),which not only fulfills transfer learning task but also has certain local learning capability.The MWME can complete traditional feature extraction as well as domain adaptation learning in two domains whose distributions are different but relative,thus indicating its better robustness and adaptation.Tests show the above-proposed advantages of the MLMD criterion and the MWME method.
作者 皋军 黄丽莉
出处 《电子学报》 EI CAS CSCD 北大核心 2013年第8期1462-1468,共7页 Acta Electronica Sinica
基金 国家自然科学基金(No.61272210) 江苏省自然科学基金(No.BK2011417) 苏州大学江苏省计算机信息处理技术重点实验室开放课题(No.KJS1126) 江苏省新型环保重点实验室开放课题(No.AE201068) 江苏省高校优秀中青年教师和校长境外研修计划
关键词 最大均值差异嵌入 最大局部均值差异 最大局部加权均值差异嵌入 特征提取 迁移学习 maximum mean discrepancy embedding maximum local weighted mean discrepancy maximum local weighted mean discrepancy embedding feature extraction transfer learning
  • 相关文献

参考文献15

  • 1边肇祺 张学工 等.模式识别[M].北京:清华大学出版社,2001..
  • 2Jolliffe I T. Principal Component Analysis[M]. New Yt~rk : Springer-Verlag, 1986.
  • 3Fislm" R A. The use of multiple ~ts in taxonomic problems[J]. Annals of Eugenics, 1936,7(2) : 179 - 188.
  • 4He X F, Niyogi P. l_x~cality preserving projections[ C/OL]. http://peples, cs. uchic ago. edtt/xiaofei/LPP_ NIPS03. pdf, 2003.
  • 5Pan S J, Yang Q. A survey on transfer learning [ J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22 (10) : 1345 - 1359.
  • 6Borgwar& K M, Gretton A, Rasch M J, Kriegel H P, Sch6olkopf B, Smola A J. Integraling slmcttued biological da- m by kernel maximum mean discrep~cy [ J]. Bioinfcrmatics, 2006,22(14) :49 - 57.
  • 7Pan J L, Kwok .1 T, Yang Q. Transfer learning via dimensional- ity reduction [ C/OL]. http://www, aaai. oig/Pape~AAAI/ 2008/AAAI0 - 108. pdf.
  • 8王雪松,潘杰,程玉虎.基于知识迁移的Ant-Q算法[J].电子学报,2011,39(10):2359-2365. 被引量:4
  • 9于重重,田蕊,谭励,涂序彦.非平衡样本分类的集成迁移学习算法[J].电子学报,2012,40(7):1358-1363. 被引量:28
  • 10Christopher G. Atkeson, Andrew W. Moore, Stefan Scb.aal. Locally weighted learning[ J]. Artifical Intelligence Review, 1997,11(1 - 5) : 11 - 73.

二级参考文献25

  • 1L M Gambardella,M Dorigo. Ant-Q: A reinforcement learning approach to the traveling salesman problem[ A]. Proceedings of12th International Conference on Machine learning[ C ]. New York: ACM Press, 1995.252- 260.
  • 2Y H Cheng,H T Feng,X S Wang.Actor-Critic learning based on adaptive importance sampling[ J]. Chinese Jourrlal of ElecIronics,2010,19(4) :583 - 588.
  • 3H M Rais, Z A Othman, A R Hamdan. Improved dynamic ant colony system (DACS) on symmetric traveling salesman prob-lem[A]. Proceedings of Intematonal Conference on Intelligent and Advanced Systems [ C ]. Piscataway: IEEE Inc, 2008.43 - 48.
  • 4N A Vien, N H Viet, S G Lee, T. H. Chung. Obstacle avoid- ance path planning for mobile robot based on Ant-Q reinforce-ment learning algodthm [ J ]. Lecture Notes in Computer Science, 2007,4491:704 - 713.
  • 5L Machado,R Schirru. The Ant-Q algorithm applied to the nuclear reload problem[ J]. International Journal of Annals of Nuclear Energy,2002,29(12) : 1455 - 1470.
  • 6C E Mariano, E Morelos. A multiple objective Ant-Q algorithm for the design of water dislribution irrigation[ A ]. Proceedings of the Genetic and Evolutionary Computation Conference[ C ]. San Francisco:Morgan Kaufmann, 1999.894 - 901.
  • 7X J Liu,Z H Ni.Ant-Q algorithm based optimization approach for process planning[ A]. Proceedings of the 8th mEE Intema-fional Conference on Control and Automation[ C]. Piscataway: mEE Inc., 2010.620 - 623.
  • 8X R Wang, T J Wu. The Ant(λ) ant colony optimization algo- rithm based on eligibility lrace [ A ]. Proceedings of the International Conference on Systems, Man and Cybernetics [ C]. Piscataway: IEEE Inc., 2003.4065 - 4070.
  • 9S G Lee,T C Chung.A reinforcement learning algorithm using temporal di? erence error in ant model [ J ]. LecRire Notes in Computer Science, 2005,3512: 217 - 224.
  • 10S J Pan, Q Yang. A survey on Iransfer learning [J]. IEE,E Transactions on Knowledge and Data Engineering, 2010, 22 (10) : 1345 - 1359.

共引文献87

同被引文献40

  • 1刘英博,王建民.面向缺陷分析的软件库挖掘方法综述[J].计算机科学,2007,34(9):1-4. 被引量:11
  • 2DAVID M J T, ROBERT P W D. Support vector data description [J]. Machine Learning, 2004, 54(1): 45-66.
  • 3WU M, YE J. A small sphere and large margin approach for novelty detection using training data with outliers[ J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 31( 11): 2088 - 2092.
  • 4XIAO Y, LIU B, CAO L, WU X, et al. Multi-sphere support vec- tor data description for oufliers detection on multi-distribution data [ C]/! Proceedings of the 2009 IEEE International Conference on Data Mining Workshops. Piscataway: IEEE Press, 2009:82 -87.
  • 5LE TTRAN D, NGUYEN P, et al. Multiple distribution data de- scription learning method for novelty detection[ C]// Proceedings of the 2011 International Joint Conference on Neural Networks. Piscat- away: IEEE Press, 2011:2321 -2326.
  • 6CHEN Y, QIAN J, SALIGRAMA V. A new one-class SVM for a- nomaly detection[ C]// Proceedings of the 2013 IEEE International Conference on Acoustics, Speech, and Signal Processing. Piscat- away: IEEE Press, 2013:3567-3571.
  • 7SCHOLKOPF B, SMOLA A J. Learning with kernels[ M]. Cam- bridge: M1T Press, 2002:227 - 250.
  • 8TSANG I W, KWOK J T, CHEUNG P M. Core vector machines: fast SVM training on very large data sets[ J]. Journal of Machine Learning Research, 2005(6): 363-392.
  • 9HOFFMANN H. Kernel PCA for novelty detection[ J]. Pattern Rec- ognition, 2007, 40(3): 863-874.
  • 10XIAO Y, WANG H, XU W, et al. L1 norm based KPCA for nov- elty detection[ J]. Pattern Recognition, 2013, 46(1) : 389 - 396.

引证文献4

二级引证文献16

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部