多示例学习算法及其应用研究
|
摘要
随着人类收集和存储数据能力的不断增长以及计算机运算能力的飞速发展，利用计算机来分析数据的要求越来越广泛和迫切，使得机器学习的重要性越来越显著。多示例学习是一种新的机器学习方法，近年来逐渐成为机器学习领域关注的一个研究热点。它有别于传统的有监督学习、无监督学习和近年来提出的半监督学习方法，被认为是一种新的学习框架。在多示例学习中，训练集由若干个具有标签的包组成，每个包含有若干个没有标签的示例。若一个包中至少有一个示例为正类，则该包被标记为正包；若一个包中所有示例都为负类，则该包被标记为负包。希望学习系统通过对训练包的学习，能够正确预测出新包的标签。由于它的训练样本的层次性表示结构，相较于平板式的单样本属性对表示，更能反映一些现实问题的逻辑结构，使得它在区分“粗标签”对象上具有独特的优势，已经得到广泛的应用，例如：药物活性分子预测，图像检索、分类和标注，文本分类，蛋白质家族预测，目录网页和链接推荐、计算机安全、计算机辅助医学诊断等领域。
本文在分析了多示例学习算法国内外研究现状和目前仍存在问题的基础上，对多示例学习算法中依赖于单个示例、包特征的构造、包特征的降维、并行算法进行研究，提出了几个多示例学习算法，并把它们应用于图像检索和分类。主要研究成果如下：
1、针对已有的多示例学习算法应用于图像检索时存在依赖于单个示例和耗时较长的缺陷，本章提出一个基于多示例学习和贝叶斯分类的图像检索方法（MIL-Bayesian）。首先，将每幅图像分割成多个区域，把图像看作多示例学习中的包，区域看作为包中的示例；其次，计算所有图像中每个区域的多样性密度（DD）函数值，提取出可能的正区域组成一个集合，使用高斯混合函数逼近估计正区域的类条件概率密度；接着，使用贝叶斯分类器为每幅图像计算一个相对于正类图像的后验概率，并根据后验概率值大小排序返回给用户；最后，经过几轮的用户相关反馈后，用户得到一组满意的图像。在Corel图像集上的实验表明，提出的方法具有好的检索精度和高的检索效率。
2、针对多示例学习中构造包特征依赖于少数示例特征的缺陷和缩窄图像的低级特征表示与高级概念之间的语义鸿沟，本章利用密度聚类获得的簇分布信息和多示例学习框架在区分歧义性对象上的特点，提出一个基于区域特征密度聚类和多示例学习的图像分类方法(DCRF-MIL)。该方法首先将每个图像分割为多个区域，将所有区域组成一个集合，在这个区域集合上，使用密度聚类算法学习到区域特征的簇分布信息；其次，将图像看作包，区域看作包中的示例，基于区域特征的簇分布信息，将包映射为簇分布空间上的一个向量作为包的特征，使得包特征带有图像区域的语义信息；最后，使用支持向量机算法，在带有包特征的训练集上训练分类器，对测试图像进行分类。在Corel图像集和MUSK分子活性预测数据集上的实验表明，DCRF-MIL算法具有分类精度高和参数易于选择等特点。
3、针对多示例学习中变换示例空间后获得的包特征的高维问题，本章提出一个基于多个子空间集成的多示例学习算法(MSEMIL)，和它的并行实现算法(P_MSEMIL)。该方法首先将多示例学习中的包，向所有示例组成的示例空间映射得到一个包特征；其次，通过融合bagging法选取训练样本子集和随机选取特征子集的方法，将训练集和测试集划分成多个子空间，在每个子空间上训练一个半监督子分类器；随后，通过集成策略合并多个子分类器的分类结果，得到一个多示例学习集成分类器。最后，在机群计算系统上，应用基于Java的分布式并行计算中间件ProActive，实现这个集成分类器的并行算法。在MUSK和Corel数据集上的实验表明，与其它同类算法相比，MSEMIL具有分类精度高、对标签噪声健壮的特点。实验还表明，P_MSEMIL具有小的计算耗时和较高的加速比等特点。
With both the continuous increase in human being’s ability in collecting and storing dataand the rapid development of computer’s computing ability, the requirement of analyzing databy computers is more popular and more urgent than before, which makes the machinelearning techniques be more and more significant. In recent years multi-instance learning(MIL), a kind of new machine learning method, has become one of research hotspots inmachine learning field. MIL is different from the traditional supervised learning, unsupervisedlearning and the recent proposed semi-supervised learning method, so it is considered to be anew learning framework. In the framework of MIL, the training set consists of a number ofbags with labels, and each bag contains a number of unlabeled instances. A bag is labeledpositive if at least one instance in it is positive, or labeled negative if and only if all of itsinstances are negative. The objective of MIL is to construct a learning system which is learnedbased on the training bags can correctly predict the label of new bags.
Because of the hierarchical representation structure of training samples, the MIL canrepresent the logical structure of some real problems more accurately than the traditionalsingle-instance-label representation can, which makes it has the unique superiority indistinguishing the so-called "ambiguity objects". Consequently, it has been widely applied tovarious applications, such as: drug activity prediction, image retrieval, image categorization,image annotation，text categorization, protein function prediction, web index page and link ofrecommendation, computer security, computer-aided diagnostics, and so on.
Based on the analysis of research status and disadvantages of these existing MILalgorithms, this paper focuses on the research of the following issues of MIL, i.e., the relyingon a single instance, construction of bag features, reduction of bag features, parallelization ofMIL algorithms and their applications to image retrieval, image classification. Some goodexperimental results have been obtained and they can be summarized as follows:
1. Aiming at the two disadvantages of existing MIL algorithms for image retrieval, i.e.the dependence on the presence of a single instance and high time-consuming, this chapterpresents an MIL and Bayesian classification based image retrieval method (MIL-Bayesian). In MIL-Bayesian, firstly, each image is divided into several regions, and the image is viewed asone bag in MIL, each region is regarded as an instance in corresponding bag. Secondly,calculate the diversity density (DD) function values of each region, and extract the possiblepositive region to compose a set, then estimate the class conditional probability densityfunction of positive regions using Gaussian mixture model. Thirdly, a Bayesian classifier isused to calculate the posterior probability of images with positive class label, and then theretrieved results are returned to user according to the posterior probability values indescending order. Finally, after several rounds of user relevance feedback, the user gets asatisfactory image. The experimental results on the Corel image set show that the proposedmethod has good retrieval precision and high retrieval efficiency.
2. In order to narrow the semantic gap between low-level visual features and high-levelsemantic concepts in image categorization, this chapter exploits the clustering informationfrom a density clustering algorithm and the characteristics of multi-instance learningframework in distinguishing ambiguous object, proposes an image categorization methodusing density clustering on region feature and multi-instance learning, termed as DCRF-MILwhich treats image classification as a multi-instance learning problem. Firstly, it divides eachimage into a number of regions, re-lines up all regions into a collection, and then uses adensity clustering algorithm to learn the potential distribution information of region featuresin the collection. Secondly, it treats image as bag and regions as instances. Based on thecluster distribution information of region features, the bag is mapped into a vector in thecluster distribution space. Finally, a support vector machine classifier is constructed to predictthe class label of the unlabeled image. The experiments on the Corel image data set andMUSK molecular activity prediction data set show DCRF-MIL algorithm has highclassification accuracy and its parameters are easy to select.
3. Aiming at the high-dimentional problem of bag features derived from transformationof instance space in MIL, this chapter proposes one multi-sub-space based ensemble MILalgorithms (MSEMIL) and its parallel version (P_MSEMIL). Firstly, this method determinedthe bag feature by mapping the bag in MIL into the instance space which is consist of allinstances; Secondly, by incorporating the bagging-based training samples selection methodand random feature subset selection method, the training set and test set is divided into multiple sub-spaces, and the semi-supervised learning is conducted in each sub-space toobtain one corresponding classifier. Consequently one ensemble classification system can beachieved by integrating the classification results of these multiple base classifiers. Finally, incluster computing systems, the P_MSEMIL is realized by using ProActive, which isJava-based distributed parallel computing middleware. The experimental results on MUSKand Corel data sets show that, compared with other similar algorithms, MSEMIL has higherclassification accuracy and better robustness to label noise. The experimental results alsoshow that the P_MSEMIL has a lower computation time, higher speed up ratio and othercharacteristics.
本文在分析了多示例学习算法国内外研究现状和目前仍存在问题的基础上，对多示例学习算法中依赖于单个示例、包特征的构造、包特征的降维、并行算法进行研究，提出了几个多示例学习算法，并把它们应用于图像检索和分类。主要研究成果如下：
1、针对已有的多示例学习算法应用于图像检索时存在依赖于单个示例和耗时较长的缺陷，本章提出一个基于多示例学习和贝叶斯分类的图像检索方法（MIL-Bayesian）。首先，将每幅图像分割成多个区域，把图像看作多示例学习中的包，区域看作为包中的示例；其次，计算所有图像中每个区域的多样性密度（DD）函数值，提取出可能的正区域组成一个集合，使用高斯混合函数逼近估计正区域的类条件概率密度；接着，使用贝叶斯分类器为每幅图像计算一个相对于正类图像的后验概率，并根据后验概率值大小排序返回给用户；最后，经过几轮的用户相关反馈后，用户得到一组满意的图像。在Corel图像集上的实验表明，提出的方法具有好的检索精度和高的检索效率。
2、针对多示例学习中构造包特征依赖于少数示例特征的缺陷和缩窄图像的低级特征表示与高级概念之间的语义鸿沟，本章利用密度聚类获得的簇分布信息和多示例学习框架在区分歧义性对象上的特点，提出一个基于区域特征密度聚类和多示例学习的图像分类方法(DCRF-MIL)。该方法首先将每个图像分割为多个区域，将所有区域组成一个集合，在这个区域集合上，使用密度聚类算法学习到区域特征的簇分布信息；其次，将图像看作包，区域看作包中的示例，基于区域特征的簇分布信息，将包映射为簇分布空间上的一个向量作为包的特征，使得包特征带有图像区域的语义信息；最后，使用支持向量机算法，在带有包特征的训练集上训练分类器，对测试图像进行分类。在Corel图像集和MUSK分子活性预测数据集上的实验表明，DCRF-MIL算法具有分类精度高和参数易于选择等特点。
3、针对多示例学习中变换示例空间后获得的包特征的高维问题，本章提出一个基于多个子空间集成的多示例学习算法(MSEMIL)，和它的并行实现算法(P_MSEMIL)。该方法首先将多示例学习中的包，向所有示例组成的示例空间映射得到一个包特征；其次，通过融合bagging法选取训练样本子集和随机选取特征子集的方法，将训练集和测试集划分成多个子空间，在每个子空间上训练一个半监督子分类器；随后，通过集成策略合并多个子分类器的分类结果，得到一个多示例学习集成分类器。最后，在机群计算系统上，应用基于Java的分布式并行计算中间件ProActive，实现这个集成分类器的并行算法。在MUSK和Corel数据集上的实验表明，与其它同类算法相比，MSEMIL具有分类精度高、对标签噪声健壮的特点。实验还表明，P_MSEMIL具有小的计算耗时和较高的加速比等特点。
With both the continuous increase in human being’s ability in collecting and storing dataand the rapid development of computer’s computing ability, the requirement of analyzing databy computers is more popular and more urgent than before, which makes the machinelearning techniques be more and more significant. In recent years multi-instance learning(MIL), a kind of new machine learning method, has become one of research hotspots inmachine learning field. MIL is different from the traditional supervised learning, unsupervisedlearning and the recent proposed semi-supervised learning method, so it is considered to be anew learning framework. In the framework of MIL, the training set consists of a number ofbags with labels, and each bag contains a number of unlabeled instances. A bag is labeledpositive if at least one instance in it is positive, or labeled negative if and only if all of itsinstances are negative. The objective of MIL is to construct a learning system which is learnedbased on the training bags can correctly predict the label of new bags.
Because of the hierarchical representation structure of training samples, the MIL canrepresent the logical structure of some real problems more accurately than the traditionalsingle-instance-label representation can, which makes it has the unique superiority indistinguishing the so-called "ambiguity objects". Consequently, it has been widely applied tovarious applications, such as: drug activity prediction, image retrieval, image categorization,image annotation，text categorization, protein function prediction, web index page and link ofrecommendation, computer security, computer-aided diagnostics, and so on.
Based on the analysis of research status and disadvantages of these existing MILalgorithms, this paper focuses on the research of the following issues of MIL, i.e., the relyingon a single instance, construction of bag features, reduction of bag features, parallelization ofMIL algorithms and their applications to image retrieval, image classification. Some goodexperimental results have been obtained and they can be summarized as follows:
1. Aiming at the two disadvantages of existing MIL algorithms for image retrieval, i.e.the dependence on the presence of a single instance and high time-consuming, this chapterpresents an MIL and Bayesian classification based image retrieval method (MIL-Bayesian). In MIL-Bayesian, firstly, each image is divided into several regions, and the image is viewed asone bag in MIL, each region is regarded as an instance in corresponding bag. Secondly,calculate the diversity density (DD) function values of each region, and extract the possiblepositive region to compose a set, then estimate the class conditional probability densityfunction of positive regions using Gaussian mixture model. Thirdly, a Bayesian classifier isused to calculate the posterior probability of images with positive class label, and then theretrieved results are returned to user according to the posterior probability values indescending order. Finally, after several rounds of user relevance feedback, the user gets asatisfactory image. The experimental results on the Corel image set show that the proposedmethod has good retrieval precision and high retrieval efficiency.
2. In order to narrow the semantic gap between low-level visual features and high-levelsemantic concepts in image categorization, this chapter exploits the clustering informationfrom a density clustering algorithm and the characteristics of multi-instance learningframework in distinguishing ambiguous object, proposes an image categorization methodusing density clustering on region feature and multi-instance learning, termed as DCRF-MILwhich treats image classification as a multi-instance learning problem. Firstly, it divides eachimage into a number of regions, re-lines up all regions into a collection, and then uses adensity clustering algorithm to learn the potential distribution information of region featuresin the collection. Secondly, it treats image as bag and regions as instances. Based on thecluster distribution information of region features, the bag is mapped into a vector in thecluster distribution space. Finally, a support vector machine classifier is constructed to predictthe class label of the unlabeled image. The experiments on the Corel image data set andMUSK molecular activity prediction data set show DCRF-MIL algorithm has highclassification accuracy and its parameters are easy to select.
3. Aiming at the high-dimentional problem of bag features derived from transformationof instance space in MIL, this chapter proposes one multi-sub-space based ensemble MILalgorithms (MSEMIL) and its parallel version (P_MSEMIL). Firstly, this method determinedthe bag feature by mapping the bag in MIL into the instance space which is consist of allinstances; Secondly, by incorporating the bagging-based training samples selection methodand random feature subset selection method, the training set and test set is divided into multiple sub-spaces, and the semi-supervised learning is conducted in each sub-space toobtain one corresponding classifier. Consequently one ensemble classification system can beachieved by integrating the classification results of these multiple base classifiers. Finally, incluster computing systems, the P_MSEMIL is realized by using ProActive, which isJava-based distributed parallel computing middleware. The experimental results on MUSKand Corel data sets show that, compared with other similar algorithms, MSEMIL has higherclassification accuracy and better robustness to label noise. The experimental results alsoshow that the P_MSEMIL has a lower computation time, higher speed up ratio and othercharacteristics.
引文
[1]Dietterich T. G., Lathrop R. H., Lozano-Pérez T., Solving the multiple-instance problemwith axis-parallel rectangles [J], Artificial Intelligence,1997,89(1-2):31-71.
[2]Goldberg, D. E.,&Holland, J. H.(1988). Genetic algorithms and machine learning.Machine learning [J],1988，3(2):95-99.
[3]Zucker J.D., Ganascia J.G., Changes of representation for efficient learning in structuraldomains [C]. In: Proceedings of the13th International Conference on MachineLearning, Bary, Italy,1996:543-551.
[4]Maron O., Learning from ambiguity [D]. PhD dissertation, Department of ElectricalEngineering and Computer Science, MIT, Jun.1998.
[5]Maron O., Lozano-Pérez T., A framework for multiple-instance learning [J]. Advances inNeural Information Processing Systems10, Cambridge, MA: MIT Press,1998,570-576.
[6]Zhang Q, Goldman S.A., EM-DD: an improved multiple-instance learning technique [J],Advances in Neural Information Processing Systems14, Cambridge, CA: MIT Press,2002:1073-1080.
[7]Zhou Z.H., Zhang M.L., Ensembles of multi-instance learners [J], Lecture Notes inArtificial Intelligence,2003,2837:492-502.
[8]Mason L, Baxter J, Bartlett P, et al. Boosting algorithms as gradient descent in functionspace [C]. Proc of Advances in Neural Information Processing Systems. DenverMariott City Center: IEEE Press,1999:132-138.
[9]Frank, E.&Xu, X. Applying propositional learning algorithms to multi-instance data [R].Technical Report06/03, Department of Computer Science, University of Waikato.2003.
[10]Deselaers, Thomas. A conditional random field for multiple-instance learning [C], In:Proceedings of the27th International Conference on Machine Learning,2010:287-294.
[11]Ray S, Page D. Multiple instance regression [C], In: Proceedings of the18th InternationalConference on13Machine Learning, Williamstown, MA,2001,425-432.
[12]Andrews S, Tsochantaridis I, Hofmann T. Support vector machines for multiple-instancelearning [C]. Proc of Advances in Neural Information Processing Systems Vancouver:MIT Press,2002:561-568.
[13]Zhou Z H, Xu J M. On the relation between multi-instance learning and semi-supervisedlearning [C]. Proc of the24th Int Conf on Machine Learning. Corvallis,2007:1167-1174.
[14]Yang Changbo, Dong Ming, Hua Jing. Region-based image annotation usingasymmetrical support vector machine-based multiple-Instance learning [C].Proceedings of the IEEE Computer Society Conference on Computer Vision andPattern Recognition,2006,2:2057-2063.
[15]Peter V Gehler, Olivier Chapelle. Deterministic annealing for multiple-instance learning
[C]. Proc of the11th Int Conf on Artificial Intelligence and Statistics. San Juan,2007:123-130.
[16]Bunescu R C, Mooney R J. Multiple instance learning for sparse positive bags [C]. Procof the24th Int Conf on Machine Learning. New York,2007:105-112.
[17]路晶,马少平.使用多例学习的启发式SVM算法的图像自动标注[J],计算机研究与发展,2009,46(5):864-871.
[18]Zhou Z-H, Zhang M-L. Neural networks for multi-instance learning [R]. Technical Report,AI Lab, CS Dept., Nanjing Univ., Aug.2002.
[19]Zhang M-L, Zhou Z-H. Improve multi-instance neural networks through featureselection[J], Neural Processing Letters,2004,19(1):1-10.
[20]J.Ramon and L.De Raedt, Multi-instance neural networks [C], Proc. ICML-2000Workshop Attribute-Value and Relational Learning,2000.
[21]张敏灵,周志华.基于神经网络的多示例回归算法[J],软件学报,2003,14:1238-1242.
[22]Ruffo G. Learning single and multiple instance decision trees for computer securityapplications [D], PhD dissertation, Department of Computer Science, University ofTurin, Torino, Italy, Feb.2000.
[23]Chevaleyre Y, Zucker J-D. Solving multiple-instance and multiple-part learning problemswith decision trees and decision rules [C], In: Proceedings of the14th BiennialConference of the Canadian Society for Computational Studies of Intelligence (LNAI2056), Ottawa, Canada,2001:204-214.
[24]Wang Jun, Zucker Jean-Daniel. Solving the multipleinstance problem: A lazy learningapproach [C]. Proc of the17th Int Conf on Machine Learning. San Francisco,2000:1119-1125.
[25]李大湘,彭进业,贺静芳.基于EMD-kNN多示例学习算法的图像分类[J].光电子激光,2010,21(2):302-306.
[26]Yossi R, Carlo T, Leonidas J G. The earthmover’s distance as a metric for image retrieval[J]. Int J of Computer Vision,2000,40(2):99-121.
[27]Srensen, Lauge Loog, Marco, et al. Dissimilarity-based multiple instance learning [J].Lecture Notes in Computer Science,2010,6218:129-138.
[28]Chen Yixin, Wang Jame Ze. Image categorization by learning and reasoning with regions[J], Joural of Machine Learning Research,2004,5(8):913-939.
[29]Chen Yixin, Bi Jinbo, Wang James Ze. MILES: Multiple instance learning via embeddedinstance selection [J]. IEEE Trans on Pattern Analysis and Machine Intelligence,2006,28(12):1931-1947.
[30]Zhang M L, Zhou Z H. Multi-instance clustering with applications to multi-instanceprediction [J]. Applied Intelligence,2009,31(1):47-68.
[31]Yajima Y., Kuo Tien-Fang, Efficient formulations for1-SVM and their application torecommendation tasks [J]. Journal of Computers,2006,1(3):27-34.
[32] Rouhollah Rahmani, Sally A Goldman. MISSL: Multiple instance semi-supervisedlearning [C]. Proc of the23rd Int Conf on Machine Learning. Pittsburgh: CarnegieMellon University Press,2006:705-712.
[33]Wang Changhu, Zhang Lei, Zhang Hongjiang, Graph-based multiple-instance learning forobject-based image retrieval [C], Proceedings of the1st International ACMConference on Multimedia Information Retrieval, MIR2008:156-163.
[34]李大湘,彭进业,李展.基于半监督多示例学习的对象图像检索[J].控制与决策,2010,25(7):981-986.
[35]李展,彭进业,温超.基于谱聚类和多示例学习的图像检索方法[J].华南理工大学（自然科学版），2011,39(7):156-162.
[36]甘睿,印鉴.通过挖掘示例中的概念来解决多示例学习问题[J].计算机研究与发展,2011,48:73-78.
[37]温超,耿国华,李展.构建新包空间的多示例学习方法[J].西安交通大学学报,2011,45(8):62-66,117.
[38]Weidmann N, Frank E, Pfahringer B. A two-level learning method for generalizedmulti-instance problems [C], In: Proceedings of the14th European Conference onMachine Learning (LNAI2837), Cavtat-Dubrovnik, Croatia,2003,468-479.
[39]Scott S., Zhang J., Brown J., On generalized multiple-instance learning [R], TechnicalReport UNL-CSE-2003-5, Department of Computer Science and Engineering,University of Nebraska, Lincoln, NE,2003.
[40]Tao Q., Scott S., A faster algorithm for generalized multiple-instance learning [C]. InProceedings of the17th International Florida Artificial Intelligence Research SocietyConference,2004:550–555.
[41]Zhou Z.-H., Sun Y.-Y., Li Y.-F., Multi-instance learning by treating instances as non-I.I.D.samples [C], In Proceedings of the26th International Conference on MachineLearning,2009:1249-1256.
[42]Herman, Gunawan Ye, Getian, et al., Multi-instance learning with relational informationof instances [C], In: Proceeding of2009Workshop on Applications of ComputerVision, WACV2009:1-7.
[43]Kim Minyoung, Gaussian processes multiple instance learning [C], In: Proceedings of27th International Conference on Machine Learning,2010:535-542.
[44]Mangasarian O.L., Wild E.W., Multiple instance classification via successive linearprogramming [J], Journal of Optimization Theory and Applications.2007,137(3):555-568.
[45]Burr Settles, Mark Craven, Soumya Ray Multiple-instance active learning [J], Journal ofIn Advances in Neural Information Processing Systems,2008:1289-1296.
[46]Sun Yu-Yin Ng, Michael K., Zhou Zhi-Hua, Multi-instance dimensionality reduction [C],In: Proceedings of the National Conference on Artificial Intelligence,2010,1:587-592.
[47]詹德川,周志华.基于流形学习的多示例回归算法[J]，计算机学报，2006,11:1948-1955.
[48]蔡自兴,李枚毅.多示例学习及其研究现状[J],控制与决策,2004,6:611-615.
[49]Zhou Z.-H., Multi-instance learning: a survey [R]. Technical report, AI Lab, Departmentof Computer Science&Technology, Nanjing University, Nanjing, China,2004.
[50]李大湘,赵小强,李娜.图像语义分析的多示例学习算法综述[J].控制与决策,2013,28(4):481-488.
[51]Foulds James, Frank Eibe, A review of multi-instance learning assumptions [J], theknowledge engineering review,2010,2:1-24.
[52]Ray S., Craven M., Supervised versus multiple instance learning: An empiricalcomparison [C], Proc.22nd Int’l Conf. Machine Learning,2005:697-704.
[53]Zhou, Zhi-Hua Multi-instance learning from supervised view [J], Journal of ComputerScience and Technology,2006,21(5):800-809.
[54]黎铭,薛晓冰,周志华,基于多示例学习的中文Web目录页面推荐[J],软件学报,2004(15),1328-1335.
[55]薛晓冰韩洁凌等，基于多示例学习技术的Web目录页面链接推荐[J],计算机研究与发展,2007(44),406-411.
[56]Zhu Liang, Zhao Bo, Gao Yang, Multi-class multi-instance learning for lung cancer imageclassification based on bag feature selection [C], In: Proceedings of the fifthinternational conference on fuzzy system and knowledge discovery.2008:487-492.
[57]Kotsiantis, S. Kanellopoulos, D. Multi-instance learning for predicting fraudulentfinancial statements [C]. In: Proceeding of the Third International Conference onConvergence and Hybrid Information Technology,2008:448-452.
[58]Kotsiantis S. Kanellopoulos D. Multi-instance learning for bankruptcy prediction [C]. In:Proceeding of the third International Conference on Convergence and HybridInformation Technology,2008:1007-1012.
[59]崔永艳.基于多示例学习的异常行为检测方法研究[D].南京:南京大学计算机科学与技术系,2011.
[60]Ruffo G. Learning single and multiple instance decision trees for computer securityapplications [D]. PhD dissertation, Department of Computer Science, University ofTurin, Torino, Italy, Feb.2000.
[61]Fliekner M., Sawhney H., et al. Query by image and video content: the QBIC System [J],IEEE Computer,1995,28(9):23-32.
[62]Pentland A., Pieard R.W., Selaroff S., Photobook: Content-based manipulation of imagedatabases [J], International Journal of Computer Vision,1996,18(3):233-254.
[63]Smith J.R., Chang S.F., Visua1SEEK: a fully automated content-based image querysystenm [C], In proeeedings of the fourth ACM international conference onMultimedia, MA, Nov.1996:87-98.
[64]Ma W.Y., Manjunath B., NETRA: A toolbox for navigating large image databases [J],Multimedia Systems,1999,7(3):184-198.
[65]Carson C., Thomas M.,Belongie S., et al., Blobworld:a system for region-based imageindexing and retrieval [C],in Proceedings of the Third International Conference onVisual Information Systems,1999,1614:509-516.
[66]Wang J.Z., Li J., Wiederhold G. SIMPLIeity: semantics-sensitive integrated matching forpicture libraries [J], IEEE Transactions on Pattern Analysis and Machine Intelligence,2001,23(9):947-963.
[67]Aggarwa G., Asllwin T.V., Ghosal S. An image retrieval system with automatic querymodifieation [J], IEEE Transactions on Multimedia,2002,4(2):201-214.
[68]Babenko B., Verma N., Dollár P., Belongie S., Multiple instance learning with manifoldbags [C], Proceedings of the28th International Conference on Machine Learning,ICML2011:81-88.
[69]He Jianjun, Gu Hong, Wang Zhelong, Bayesian multi-instance multi-label learning usingGaussian process prior [J], Machine Learning,2012,88(1-2):273-295.
[70]Maron O. Learning from ambiguity [D]. PhD dissertation, Department of ElectricalEngineering and Computer Science, MIT, Jun.1998.
[71]Yang C, Lozano-P erez T. Image database retrieval with multiple-instance learningtechniques [C]. In: Proceedings of the16th International Conference on DataEngineering, San Diego, CA,2000:233-243.
[72]Zhang Q., Yu W., Goldman S.A., Fritts J.E., Content-based image retrieval usingmultiple-instance learning [C]. In: Proceedings of the19th International Conferenceon Machine Learning, Sydney, Australia,2002:682-689.
[73]Zhou Zhi-Hua, Zhang Min-Ling, Chen Ke-Jia, A novel bag generator for image databaseretrieval with multi-instance learning Techniques [C], In: Proceeding of the15th IEEEInter national Conference on Tools with Artificial Intelligence (ICTAI'2003),2003:233-238.
[74]Unser M., Texture classification and segmentation using wavelet frames [J]. IEEETransactions on Image Processing,1995,4(11):1549–1560.
[75]Li Yu-feng, Kwok James T., Tsang Ivorw, Zhou Zhi-hua, A convex method for locatingregions of interest with multi-instance learning [J]. Lecture Notes in ComputerScience,2009,5782(2):15-30.
[76]Zhou Z.H., Zhang M.L., Solving multi-instance problems with classifier ensemble basedon constructive clustering [J]. Knowledge and Information Systems,2007,11(2):155-170.
[77]戴宏斌,张敏灵,周志华.一种基于多示例学习的图像检索方法[J],模式识别与人工智能,2006(19),179-185.
[78]Weidmann N, Frank E, Pfahringer B., A two-level learning method for generalizedmulti-instance problems [C], In: Proceedings of the14th European Conference onMachine Learning (LNAI2837), Cavtat-Dubrovnik, Croatia,2003:468-479.
[79]彭瑜,乔奇峰,魏昆娟.基于多示例学习的图像检索方法[J],中文信息学报,2008(22):64-69.
[80]Hu Xiaohong, Qian Xu, et al. A novel region-based image annotation using multi-instancelearning [C]. In: Proceedings of the Second International Workshop on knowledgeDiscovery and Data Mining,2009:602-605.
[81]Zhou Z-H, Zhang M-L., Multi-instance multi-label learning with application to sceneclassification [J]. Advances in Neural Information Processing Systems19(NIPS’06),Cambridge, MA: MIT Press,2007,1609-1616.
[82]李大湘,彭进业,李展.基于半监督多示例学习的对象图像检索[J],控制与决策,2010（7）,981-986.
[83]Li Daxiang, Peng Jinye, Object-based image retrieval using semi-supervisedmulti-instance learning [C], In: Proceedings of the20092nd International Congress onImage and Signal Processing,2009:1-5.
[84]Wang H., Huang H., Kamangar F., Maximum Margin Multi-Instance Learning [C].Proceedings of the24th Neural Information Processing Systems (NIPS,2011).Vancouver, Canada: MIT Press,2011:153-1613.
[85]曾璞,面向语义提取的图像分类关键技术研究[D],长沙,国防科技大学,2009.
[86]夏定元,基于内容的图像检索通用技术研究及应用[D].武汉：华中科技大学博士学位论文,2004.
[87]程起敏,基于内容的遥感图像库检索关键技术研究[D].北京:中国科学院遥感应用研究所博士学位论文,2004.
[88]He Xiaofei, Cai Deng, Learning a maximum margin Subspace for image retrieval [J],IEEE Transaction on Knowledge and Data Engineering,2008,20(2):189-201.
[89]Feng Songhe, Xu De, Transductive Multi-instance multi-label learning algorithm withapplication to automatic image annotation [J], Expert Systems with Applications,2010,37:661-670.
[90]Li Daxiang, Fan Jiulun, et al. Latent topic based multi-instance learing method forlocalized content-based image retrieval [J], Computers and Mathematics withApplications,2012,64:500-510.
[91]Ko Byoung Chul, Lee JiHyeon, Nam Jae-Yeal, Automatic medical image annotation andkeyword-based image retrieval using relevance feedback [J], Journal of DigitalImaging,2012,25(4):454-465.
[92]Binder A., Mueller K.R., Kawanabe M., On taxonomies for multi-class imagecategorization [J], International Journal of Computer Vision,2012,99(3):281-301.
[93]Hu Xiaohong, Qian Xu et al., A novel region-based image annotation using multi-instancelearning [C], Proceedings of the Second International Workshop on knowledgeDiscovery and Data Mining,2009:602-605.
[94]Song Xiangfa, Jiao L.C., Yang Shuyuan, Sparse coding and classifier ensemble basedmulti-instance learning for image categorization [J], Signal Processing,2013,93(1):1-11.
[95]Fu Zhouyu, Robles-Kelly A., Zhou Jun, Multiple instance learning with instance selection[J], IEEE Transactions on Pattern Analysis and Machine Intelligence,2011,33(5):958-977.
[96]Zhou Zhi-Hua, Multi-instance learning from supervised view [J]. Journal of ComputerScience and Technology,2006,6(5):800-809.
[97]Wang J. Z., Li J., Wiederhold G.，SIMPLIcity: Semantics-sensitive integrated matchingfor picture libraries [J]， IEEE Transactions on Pattern Analysis and MachineIntelligence,2001，23(9):947–963.
[98]Yang Christopher C., Ng Tobun Dorbin, Analyzing and visualizing web opiniondevelopment and social interactions with density-based clustering [J], IEEETransactions on Systems, Man, and Cybernetics Part A: Systems and Humans,2011,41(6):1144-1155.
[99]Kundakcioglu, OE, Seref O, Pardalos PM, Multiple instance learning via marginmaximization [J], Applied Numerical Mathematics,2010,60(4):14-23.
[100]Yang Jun, Jiang Yu-Gang, et al., Evaluating bag-of-visual-words representations in sceneclassification [C], Proceedings of the9th ACM SIG Multimedia InternationalWorkshop on Multimedia Information Retrieval,2007:197-206.
[101]Tao Dacheng, Asymmetric bagging and random subspace for support vectormachines-based relevance feedback in image retrieval [J], IEEE Transactions onPattern Analysis and Machine Intelligence,2006,28(7):1088-1099
[102]He Xiaofei, Cai Deng, Learning a maximum margin subspace for image retrieval [J],2008,20(2):189:201.
[103]Xu Dong, Yan shuicheng, semi-supervised bilinear subspace learning [J], IEEETransaction on Image Processing,2009,18(7):1671-1676.
[104]Ho T.K., The random subspace method for constructing decision forests [J], IEEETransactions on Pattern Analysis and Machine Intelligence,1998,20(8):832-844.
[105]杨明,王飞.一种基于局部随机子空间的分类集成算法[J].模式识别与人工智能,2012,25(4):595-603.
[106]王娇,罗四维,曾宪华.基于随机子空间的半监督协同训练算法[J].电子学报,2008,36(12):60-65.
[107]苏艳,居胜峰,王中卿等.基于随机特征子空间的半监督情感分类方法研究[J].中文信息学报,2012,26(4):85-90.
[108]Yu Guoxian, Zhang Guoji, Carlotta D., Semi-Supervised classification based on randomsubspace dimensionality reduction [J]. Pattern Recognition,2012,25(3):119-1135.
[109]Yu Guoxian, Zhang Guoji, et al., Semi-Supervised ensemble classification in subspaces[J]. Applied Soft Computing,2012,12:1511-1522.
[110]余国先，张国基等，一种基于多图的集成直推分类方法[J],电子信息学报，2011，33（8）：1883-1888.
[111]Shahani B., Landgrebe D., The effect of unlabeled samples in reducing the small samplesize problem and mitigating the huges phenomenon [J]. IEEE Transactions onGeoscience and Remote Sensing,1994,32(5):1087-1095.
[112]Lippmann R.P. Pattern classification using neural networks [J]. IEEE Communications,1989,27(11):47-64.
[113]Miller D.J., Uyar H.S., A mixture of experts classifier with learning based on bothlabeled and unlabelled data [J], Advances in Neural Information Processing Systems,1997:571-577
[114]Joachims T. Transductive inference for text classification using support vector machines
[C]. In Proceedings of the16th International Conference on Machine Learning,1999:200-209.
[115]Lawrence N.D., Jordan M.I. Semi-supervised learning via Gaussian processes [J],Advances in Neural Information Processing Systems,2005:753-760.
[116]Grandvalet Y., Bengio Y., Semi-supervised learning by entropy minimization [J],Advances in Neural Information Processing Systems,2005:529-536.
[117]Zhu X., Ghahramani Z., Lafferty J., Semi-supervised learning using Gaussian fields andharmonic functions [C], In Proceedings of the20th International Conference onMachine Learning,2003:912-919.
[118]Zhou D., Bousquet O., Lal T.N., et al., Learning with local and global consistency [J],Advances in Neural Information Processing Systems,2004:321-328.
[119]Nigam K., McCallum A.K., et al., Text classification from labeled and unlabeleddocuments using EM [J]. Machine Learning,2000,39(2-3):103-134.
[120]Blum A., Chawla S., Learning from labeled and unlabeled data using graph mincuts [C],In Proceedings of the18th International Conference on Machine Learning,2001:19-26.
[121]Beldin M. Niyogi P., Sinwani V., On manifold regularization [C], In Proceedings of the10th International Workshop on Artificial Intelligence and Statistics,2005:17-24.
[122]Chen Tao, Liang Zhengyou, A distributed parallel algorithm for web page invertedindexes construction on the cluster computing systems[C]，Proceedings of2009International Forum on Information Technology and Applications,2009,2:33-36.
[123]Chen Tao, Deng Huifang, Optimal strategy of web page text allocation for invertedindex construction on heterogeneous cluster computing system [J], Journal ofInformation&Computational Science,2011,8(8):1383-1388.
[2]Goldberg, D. E.,&Holland, J. H.(1988). Genetic algorithms and machine learning.Machine learning [J],1988，3(2):95-99.
[3]Zucker J.D., Ganascia J.G., Changes of representation for efficient learning in structuraldomains [C]. In: Proceedings of the13th International Conference on MachineLearning, Bary, Italy,1996:543-551.
[4]Maron O., Learning from ambiguity [D]. PhD dissertation, Department of ElectricalEngineering and Computer Science, MIT, Jun.1998.
[5]Maron O., Lozano-Pérez T., A framework for multiple-instance learning [J]. Advances inNeural Information Processing Systems10, Cambridge, MA: MIT Press,1998,570-576.
[6]Zhang Q, Goldman S.A., EM-DD: an improved multiple-instance learning technique [J],Advances in Neural Information Processing Systems14, Cambridge, CA: MIT Press,2002:1073-1080.
[7]Zhou Z.H., Zhang M.L., Ensembles of multi-instance learners [J], Lecture Notes inArtificial Intelligence,2003,2837:492-502.
[8]Mason L, Baxter J, Bartlett P, et al. Boosting algorithms as gradient descent in functionspace [C]. Proc of Advances in Neural Information Processing Systems. DenverMariott City Center: IEEE Press,1999:132-138.
[9]Frank, E.&Xu, X. Applying propositional learning algorithms to multi-instance data [R].Technical Report06/03, Department of Computer Science, University of Waikato.2003.
[10]Deselaers, Thomas. A conditional random field for multiple-instance learning [C], In:Proceedings of the27th International Conference on Machine Learning,2010:287-294.
[11]Ray S, Page D. Multiple instance regression [C], In: Proceedings of the18th InternationalConference on13Machine Learning, Williamstown, MA,2001,425-432.
[12]Andrews S, Tsochantaridis I, Hofmann T. Support vector machines for multiple-instancelearning [C]. Proc of Advances in Neural Information Processing Systems Vancouver:MIT Press,2002:561-568.
[13]Zhou Z H, Xu J M. On the relation between multi-instance learning and semi-supervisedlearning [C]. Proc of the24th Int Conf on Machine Learning. Corvallis,2007:1167-1174.
[14]Yang Changbo, Dong Ming, Hua Jing. Region-based image annotation usingasymmetrical support vector machine-based multiple-Instance learning [C].Proceedings of the IEEE Computer Society Conference on Computer Vision andPattern Recognition,2006,2:2057-2063.
[15]Peter V Gehler, Olivier Chapelle. Deterministic annealing for multiple-instance learning
[C]. Proc of the11th Int Conf on Artificial Intelligence and Statistics. San Juan,2007:123-130.
[16]Bunescu R C, Mooney R J. Multiple instance learning for sparse positive bags [C]. Procof the24th Int Conf on Machine Learning. New York,2007:105-112.
[17]路晶,马少平.使用多例学习的启发式SVM算法的图像自动标注[J],计算机研究与发展,2009,46(5):864-871.
[18]Zhou Z-H, Zhang M-L. Neural networks for multi-instance learning [R]. Technical Report,AI Lab, CS Dept., Nanjing Univ., Aug.2002.
[19]Zhang M-L, Zhou Z-H. Improve multi-instance neural networks through featureselection[J], Neural Processing Letters,2004,19(1):1-10.
[20]J.Ramon and L.De Raedt, Multi-instance neural networks [C], Proc. ICML-2000Workshop Attribute-Value and Relational Learning,2000.
[21]张敏灵,周志华.基于神经网络的多示例回归算法[J],软件学报,2003,14:1238-1242.
[22]Ruffo G. Learning single and multiple instance decision trees for computer securityapplications [D], PhD dissertation, Department of Computer Science, University ofTurin, Torino, Italy, Feb.2000.
[23]Chevaleyre Y, Zucker J-D. Solving multiple-instance and multiple-part learning problemswith decision trees and decision rules [C], In: Proceedings of the14th BiennialConference of the Canadian Society for Computational Studies of Intelligence (LNAI2056), Ottawa, Canada,2001:204-214.
[24]Wang Jun, Zucker Jean-Daniel. Solving the multipleinstance problem: A lazy learningapproach [C]. Proc of the17th Int Conf on Machine Learning. San Francisco,2000:1119-1125.
[25]李大湘,彭进业,贺静芳.基于EMD-kNN多示例学习算法的图像分类[J].光电子激光,2010,21(2):302-306.
[26]Yossi R, Carlo T, Leonidas J G. The earthmover’s distance as a metric for image retrieval[J]. Int J of Computer Vision,2000,40(2):99-121.
[27]Srensen, Lauge Loog, Marco, et al. Dissimilarity-based multiple instance learning [J].Lecture Notes in Computer Science,2010,6218:129-138.
[28]Chen Yixin, Wang Jame Ze. Image categorization by learning and reasoning with regions[J], Joural of Machine Learning Research,2004,5(8):913-939.
[29]Chen Yixin, Bi Jinbo, Wang James Ze. MILES: Multiple instance learning via embeddedinstance selection [J]. IEEE Trans on Pattern Analysis and Machine Intelligence,2006,28(12):1931-1947.
[30]Zhang M L, Zhou Z H. Multi-instance clustering with applications to multi-instanceprediction [J]. Applied Intelligence,2009,31(1):47-68.
[31]Yajima Y., Kuo Tien-Fang, Efficient formulations for1-SVM and their application torecommendation tasks [J]. Journal of Computers,2006,1(3):27-34.
[32] Rouhollah Rahmani, Sally A Goldman. MISSL: Multiple instance semi-supervisedlearning [C]. Proc of the23rd Int Conf on Machine Learning. Pittsburgh: CarnegieMellon University Press,2006:705-712.
[33]Wang Changhu, Zhang Lei, Zhang Hongjiang, Graph-based multiple-instance learning forobject-based image retrieval [C], Proceedings of the1st International ACMConference on Multimedia Information Retrieval, MIR2008:156-163.
[34]李大湘,彭进业,李展.基于半监督多示例学习的对象图像检索[J].控制与决策,2010,25(7):981-986.
[35]李展,彭进业,温超.基于谱聚类和多示例学习的图像检索方法[J].华南理工大学（自然科学版），2011,39(7):156-162.
[36]甘睿,印鉴.通过挖掘示例中的概念来解决多示例学习问题[J].计算机研究与发展,2011,48:73-78.
[37]温超,耿国华,李展.构建新包空间的多示例学习方法[J].西安交通大学学报,2011,45(8):62-66,117.
[38]Weidmann N, Frank E, Pfahringer B. A two-level learning method for generalizedmulti-instance problems [C], In: Proceedings of the14th European Conference onMachine Learning (LNAI2837), Cavtat-Dubrovnik, Croatia,2003,468-479.
[39]Scott S., Zhang J., Brown J., On generalized multiple-instance learning [R], TechnicalReport UNL-CSE-2003-5, Department of Computer Science and Engineering,University of Nebraska, Lincoln, NE,2003.
[40]Tao Q., Scott S., A faster algorithm for generalized multiple-instance learning [C]. InProceedings of the17th International Florida Artificial Intelligence Research SocietyConference,2004:550–555.
[41]Zhou Z.-H., Sun Y.-Y., Li Y.-F., Multi-instance learning by treating instances as non-I.I.D.samples [C], In Proceedings of the26th International Conference on MachineLearning,2009:1249-1256.
[42]Herman, Gunawan Ye, Getian, et al., Multi-instance learning with relational informationof instances [C], In: Proceeding of2009Workshop on Applications of ComputerVision, WACV2009:1-7.
[43]Kim Minyoung, Gaussian processes multiple instance learning [C], In: Proceedings of27th International Conference on Machine Learning,2010:535-542.
[44]Mangasarian O.L., Wild E.W., Multiple instance classification via successive linearprogramming [J], Journal of Optimization Theory and Applications.2007,137(3):555-568.
[45]Burr Settles, Mark Craven, Soumya Ray Multiple-instance active learning [J], Journal ofIn Advances in Neural Information Processing Systems,2008:1289-1296.
[46]Sun Yu-Yin Ng, Michael K., Zhou Zhi-Hua, Multi-instance dimensionality reduction [C],In: Proceedings of the National Conference on Artificial Intelligence,2010,1:587-592.
[47]詹德川,周志华.基于流形学习的多示例回归算法[J]，计算机学报，2006,11:1948-1955.
[48]蔡自兴,李枚毅.多示例学习及其研究现状[J],控制与决策,2004,6:611-615.
[49]Zhou Z.-H., Multi-instance learning: a survey [R]. Technical report, AI Lab, Departmentof Computer Science&Technology, Nanjing University, Nanjing, China,2004.
[50]李大湘,赵小强,李娜.图像语义分析的多示例学习算法综述[J].控制与决策,2013,28(4):481-488.
[51]Foulds James, Frank Eibe, A review of multi-instance learning assumptions [J], theknowledge engineering review,2010,2:1-24.
[52]Ray S., Craven M., Supervised versus multiple instance learning: An empiricalcomparison [C], Proc.22nd Int’l Conf. Machine Learning,2005:697-704.
[53]Zhou, Zhi-Hua Multi-instance learning from supervised view [J], Journal of ComputerScience and Technology,2006,21(5):800-809.
[54]黎铭,薛晓冰,周志华,基于多示例学习的中文Web目录页面推荐[J],软件学报,2004(15),1328-1335.
[55]薛晓冰韩洁凌等，基于多示例学习技术的Web目录页面链接推荐[J],计算机研究与发展,2007(44),406-411.
[56]Zhu Liang, Zhao Bo, Gao Yang, Multi-class multi-instance learning for lung cancer imageclassification based on bag feature selection [C], In: Proceedings of the fifthinternational conference on fuzzy system and knowledge discovery.2008:487-492.
[57]Kotsiantis, S. Kanellopoulos, D. Multi-instance learning for predicting fraudulentfinancial statements [C]. In: Proceeding of the Third International Conference onConvergence and Hybrid Information Technology,2008:448-452.
[58]Kotsiantis S. Kanellopoulos D. Multi-instance learning for bankruptcy prediction [C]. In:Proceeding of the third International Conference on Convergence and HybridInformation Technology,2008:1007-1012.
[59]崔永艳.基于多示例学习的异常行为检测方法研究[D].南京:南京大学计算机科学与技术系,2011.
[60]Ruffo G. Learning single and multiple instance decision trees for computer securityapplications [D]. PhD dissertation, Department of Computer Science, University ofTurin, Torino, Italy, Feb.2000.
[61]Fliekner M., Sawhney H., et al. Query by image and video content: the QBIC System [J],IEEE Computer,1995,28(9):23-32.
[62]Pentland A., Pieard R.W., Selaroff S., Photobook: Content-based manipulation of imagedatabases [J], International Journal of Computer Vision,1996,18(3):233-254.
[63]Smith J.R., Chang S.F., Visua1SEEK: a fully automated content-based image querysystenm [C], In proeeedings of the fourth ACM international conference onMultimedia, MA, Nov.1996:87-98.
[64]Ma W.Y., Manjunath B., NETRA: A toolbox for navigating large image databases [J],Multimedia Systems,1999,7(3):184-198.
[65]Carson C., Thomas M.,Belongie S., et al., Blobworld:a system for region-based imageindexing and retrieval [C],in Proceedings of the Third International Conference onVisual Information Systems,1999,1614:509-516.
[66]Wang J.Z., Li J., Wiederhold G. SIMPLIeity: semantics-sensitive integrated matching forpicture libraries [J], IEEE Transactions on Pattern Analysis and Machine Intelligence,2001,23(9):947-963.
[67]Aggarwa G., Asllwin T.V., Ghosal S. An image retrieval system with automatic querymodifieation [J], IEEE Transactions on Multimedia,2002,4(2):201-214.
[68]Babenko B., Verma N., Dollár P., Belongie S., Multiple instance learning with manifoldbags [C], Proceedings of the28th International Conference on Machine Learning,ICML2011:81-88.
[69]He Jianjun, Gu Hong, Wang Zhelong, Bayesian multi-instance multi-label learning usingGaussian process prior [J], Machine Learning,2012,88(1-2):273-295.
[70]Maron O. Learning from ambiguity [D]. PhD dissertation, Department of ElectricalEngineering and Computer Science, MIT, Jun.1998.
[71]Yang C, Lozano-P erez T. Image database retrieval with multiple-instance learningtechniques [C]. In: Proceedings of the16th International Conference on DataEngineering, San Diego, CA,2000:233-243.
[72]Zhang Q., Yu W., Goldman S.A., Fritts J.E., Content-based image retrieval usingmultiple-instance learning [C]. In: Proceedings of the19th International Conferenceon Machine Learning, Sydney, Australia,2002:682-689.
[73]Zhou Zhi-Hua, Zhang Min-Ling, Chen Ke-Jia, A novel bag generator for image databaseretrieval with multi-instance learning Techniques [C], In: Proceeding of the15th IEEEInter national Conference on Tools with Artificial Intelligence (ICTAI'2003),2003:233-238.
[74]Unser M., Texture classification and segmentation using wavelet frames [J]. IEEETransactions on Image Processing,1995,4(11):1549–1560.
[75]Li Yu-feng, Kwok James T., Tsang Ivorw, Zhou Zhi-hua, A convex method for locatingregions of interest with multi-instance learning [J]. Lecture Notes in ComputerScience,2009,5782(2):15-30.
[76]Zhou Z.H., Zhang M.L., Solving multi-instance problems with classifier ensemble basedon constructive clustering [J]. Knowledge and Information Systems,2007,11(2):155-170.
[77]戴宏斌,张敏灵,周志华.一种基于多示例学习的图像检索方法[J],模式识别与人工智能,2006(19),179-185.
[78]Weidmann N, Frank E, Pfahringer B., A two-level learning method for generalizedmulti-instance problems [C], In: Proceedings of the14th European Conference onMachine Learning (LNAI2837), Cavtat-Dubrovnik, Croatia,2003:468-479.
[79]彭瑜,乔奇峰,魏昆娟.基于多示例学习的图像检索方法[J],中文信息学报,2008(22):64-69.
[80]Hu Xiaohong, Qian Xu, et al. A novel region-based image annotation using multi-instancelearning [C]. In: Proceedings of the Second International Workshop on knowledgeDiscovery and Data Mining,2009:602-605.
[81]Zhou Z-H, Zhang M-L., Multi-instance multi-label learning with application to sceneclassification [J]. Advances in Neural Information Processing Systems19(NIPS’06),Cambridge, MA: MIT Press,2007,1609-1616.
[82]李大湘,彭进业,李展.基于半监督多示例学习的对象图像检索[J],控制与决策,2010（7）,981-986.
[83]Li Daxiang, Peng Jinye, Object-based image retrieval using semi-supervisedmulti-instance learning [C], In: Proceedings of the20092nd International Congress onImage and Signal Processing,2009:1-5.
[84]Wang H., Huang H., Kamangar F., Maximum Margin Multi-Instance Learning [C].Proceedings of the24th Neural Information Processing Systems (NIPS,2011).Vancouver, Canada: MIT Press,2011:153-1613.
[85]曾璞,面向语义提取的图像分类关键技术研究[D],长沙,国防科技大学,2009.
[86]夏定元,基于内容的图像检索通用技术研究及应用[D].武汉：华中科技大学博士学位论文,2004.
[87]程起敏,基于内容的遥感图像库检索关键技术研究[D].北京:中国科学院遥感应用研究所博士学位论文,2004.
[88]He Xiaofei, Cai Deng, Learning a maximum margin Subspace for image retrieval [J],IEEE Transaction on Knowledge and Data Engineering,2008,20(2):189-201.
[89]Feng Songhe, Xu De, Transductive Multi-instance multi-label learning algorithm withapplication to automatic image annotation [J], Expert Systems with Applications,2010,37:661-670.
[90]Li Daxiang, Fan Jiulun, et al. Latent topic based multi-instance learing method forlocalized content-based image retrieval [J], Computers and Mathematics withApplications,2012,64:500-510.
[91]Ko Byoung Chul, Lee JiHyeon, Nam Jae-Yeal, Automatic medical image annotation andkeyword-based image retrieval using relevance feedback [J], Journal of DigitalImaging,2012,25(4):454-465.
[92]Binder A., Mueller K.R., Kawanabe M., On taxonomies for multi-class imagecategorization [J], International Journal of Computer Vision,2012,99(3):281-301.
[93]Hu Xiaohong, Qian Xu et al., A novel region-based image annotation using multi-instancelearning [C], Proceedings of the Second International Workshop on knowledgeDiscovery and Data Mining,2009:602-605.
[94]Song Xiangfa, Jiao L.C., Yang Shuyuan, Sparse coding and classifier ensemble basedmulti-instance learning for image categorization [J], Signal Processing,2013,93(1):1-11.
[95]Fu Zhouyu, Robles-Kelly A., Zhou Jun, Multiple instance learning with instance selection[J], IEEE Transactions on Pattern Analysis and Machine Intelligence,2011,33(5):958-977.
[96]Zhou Zhi-Hua, Multi-instance learning from supervised view [J]. Journal of ComputerScience and Technology,2006,6(5):800-809.
[97]Wang J. Z., Li J., Wiederhold G.，SIMPLIcity: Semantics-sensitive integrated matchingfor picture libraries [J]， IEEE Transactions on Pattern Analysis and MachineIntelligence,2001，23(9):947–963.
[98]Yang Christopher C., Ng Tobun Dorbin, Analyzing and visualizing web opiniondevelopment and social interactions with density-based clustering [J], IEEETransactions on Systems, Man, and Cybernetics Part A: Systems and Humans,2011,41(6):1144-1155.
[99]Kundakcioglu, OE, Seref O, Pardalos PM, Multiple instance learning via marginmaximization [J], Applied Numerical Mathematics,2010,60(4):14-23.
[100]Yang Jun, Jiang Yu-Gang, et al., Evaluating bag-of-visual-words representations in sceneclassification [C], Proceedings of the9th ACM SIG Multimedia InternationalWorkshop on Multimedia Information Retrieval,2007:197-206.
[101]Tao Dacheng, Asymmetric bagging and random subspace for support vectormachines-based relevance feedback in image retrieval [J], IEEE Transactions onPattern Analysis and Machine Intelligence,2006,28(7):1088-1099
[102]He Xiaofei, Cai Deng, Learning a maximum margin subspace for image retrieval [J],2008,20(2):189:201.
[103]Xu Dong, Yan shuicheng, semi-supervised bilinear subspace learning [J], IEEETransaction on Image Processing,2009,18(7):1671-1676.
[104]Ho T.K., The random subspace method for constructing decision forests [J], IEEETransactions on Pattern Analysis and Machine Intelligence,1998,20(8):832-844.
[105]杨明,王飞.一种基于局部随机子空间的分类集成算法[J].模式识别与人工智能,2012,25(4):595-603.
[106]王娇,罗四维,曾宪华.基于随机子空间的半监督协同训练算法[J].电子学报,2008,36(12):60-65.
[107]苏艳,居胜峰,王中卿等.基于随机特征子空间的半监督情感分类方法研究[J].中文信息学报,2012,26(4):85-90.
[108]Yu Guoxian, Zhang Guoji, Carlotta D., Semi-Supervised classification based on randomsubspace dimensionality reduction [J]. Pattern Recognition,2012,25(3):119-1135.
[109]Yu Guoxian, Zhang Guoji, et al., Semi-Supervised ensemble classification in subspaces[J]. Applied Soft Computing,2012,12:1511-1522.
[110]余国先，张国基等，一种基于多图的集成直推分类方法[J],电子信息学报，2011，33（8）：1883-1888.
[111]Shahani B., Landgrebe D., The effect of unlabeled samples in reducing the small samplesize problem and mitigating the huges phenomenon [J]. IEEE Transactions onGeoscience and Remote Sensing,1994,32(5):1087-1095.
[112]Lippmann R.P. Pattern classification using neural networks [J]. IEEE Communications,1989,27(11):47-64.
[113]Miller D.J., Uyar H.S., A mixture of experts classifier with learning based on bothlabeled and unlabelled data [J], Advances in Neural Information Processing Systems,1997:571-577
[114]Joachims T. Transductive inference for text classification using support vector machines
[C]. In Proceedings of the16th International Conference on Machine Learning,1999:200-209.
[115]Lawrence N.D., Jordan M.I. Semi-supervised learning via Gaussian processes [J],Advances in Neural Information Processing Systems,2005:753-760.
[116]Grandvalet Y., Bengio Y., Semi-supervised learning by entropy minimization [J],Advances in Neural Information Processing Systems,2005:529-536.
[117]Zhu X., Ghahramani Z., Lafferty J., Semi-supervised learning using Gaussian fields andharmonic functions [C], In Proceedings of the20th International Conference onMachine Learning,2003:912-919.
[118]Zhou D., Bousquet O., Lal T.N., et al., Learning with local and global consistency [J],Advances in Neural Information Processing Systems,2004:321-328.
[119]Nigam K., McCallum A.K., et al., Text classification from labeled and unlabeleddocuments using EM [J]. Machine Learning,2000,39(2-3):103-134.
[120]Blum A., Chawla S., Learning from labeled and unlabeled data using graph mincuts [C],In Proceedings of the18th International Conference on Machine Learning,2001:19-26.
[121]Beldin M. Niyogi P., Sinwani V., On manifold regularization [C], In Proceedings of the10th International Workshop on Artificial Intelligence and Statistics,2005:17-24.
[122]Chen Tao, Liang Zhengyou, A distributed parallel algorithm for web page invertedindexes construction on the cluster computing systems[C]，Proceedings of2009International Forum on Information Technology and Applications,2009,2:33-36.
[123]Chen Tao, Deng Huifang, Optimal strategy of web page text allocation for invertedindex construction on heterogeneous cluster computing system [J], Journal ofInformation&Computational Science,2011,8(8):1383-1388.