一种基于示例非独立同分布的多示例多标签分类算法

doi:10.11896/j.issn.1002-137X.2016.02.060

摘要/Abstract

摘要： 多示例多标签学习是一种新型的机器学习框架。在多示例多标签学习中,样本以包的形式存在,一个包由多个示例组成,并被标记多个标签。以往的多示例多标签学习研究中,通常认为包中的示例是独立同分布的,但这个假设在实际应用中是很难保证的。为了利用包中示例的相关性特征,提出了一种基于示例非独立同分布的多示例多标签分类算法。该算法首先通过建立相关性矩阵表示出包内示例的相关关系,每个多示例包由一个相关性矩阵表示；然后建立基于不同尺度的相关性矩阵的核函数；最后考虑到不同标签的预测对应不同的核函数,引入多核学习构造并训练针对不同标签预测的多核SVM分类器。图像和文本数据集上的实验结果表明,该算法大大提高了多标签分类的准确性。

关键词: 多示例学习,多示例多标签学习,示例非独立同分布,多核学习

Abstract: Multi-instance multi-label learning (MIML) is a new machine learning framework.In this framework,an object is represented as a bag which is decomposed of multiple instances and labeled with multiple labels.Previous studies on MIML typically treated instances in the bags are independently identically distributed.However,it is difficult to be guaranteed in real tasks.Considering correlation features of instance in a bag,a multi-instance multi-label learning algorithm by treating instances as non-independent identically distributed samples was proposed.Firstly,instance correlations are considered by establishing an affinity matrix.By this means each bag is represented with an affinity matrix.Then,kernel functions based on the affinity matrix in different scales are established.Finally,considering predictions of different kinds of labels are corresponding to different kernels,multiple kernel learning is introduced to construct and train the MKSVMs.Experimental results on image data set and text data set show that the proposed algorithm greatly improves the accuracy of the image multi-label classification compared with other methods.

Key words: Multi-instance learning,Multi-instance multi-label learning,Non-I.I.D.instances,Multiple kernel learning

陈彤彤,丁昕苗,柳婵娟,邹海林,周树森,刘影. 一种基于示例非独立同分布的多示例多标签分类算法[J]. 计算机科学, 2016, 43(2): 287-292. https://doi.org/10.11896/j.issn.1002-137X.2016.02.060

CHEN Tong-tong, DING Xin-miao, LIU Chan-juan, ZOU Hai-lin, ZHOU Shu-sen and LIU Ying. Multi-instance Multi-label Learning Algorithm by Treating Instances as Non-independent Identically Distributed Samples[J]. Computer Science, 2016, 43(2): 287-292. https://doi.org/10.11896/j.issn.1002-137X.2016.02.060

参考文献

[1] Dietterich T G,Lathrop R H,Lozano-Pérez T.Solving the multiple instance problem with axis-parallel rectangles [J].Artificial Intelligence,1997,89(1/2):31-71
[2] Zhang Min-ling,Zhou Zhi-hua.A multi-instance regression algorithm based on neural network [J].Journal of Software,2003,14(7):1238-1242(in Chinese) 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(7):1238-1242
[3] Dai Hong-bin,Zhang Min-ling,Zhou Zhi-hua.A Multi-instance learning based approach to image retrieval [J].Pattern Recognition and Artificial Intelligence,2006,19(2):179-185(in Chinese) 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(2):179-185
[4] Li Da-xiang,Peng Jin-ye,Li Zhan.Object-based image retrieval using semi-supervised multi-instance learning algorithm [J].Control and Decision,2010,25(7):981-986(in Chinese) 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
[5] Li Da-xiang,Peng Jin-ye,Bu Qi-rong.QPSO-based multi-in-stance learning for image annotation [J].Computer Science,2010,37(6):278-282(in Chinese)李大湘,彭进业,卜起荣.基于QPSO-MIL算法的图像标注 [J].计算机科学,2010,37(6):278-282
[6] Chen Yi-xin,Bi Jin-bo,Wang J Z.MILES:Multiple-instancelearning via embedded instance selection [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2006,28(12):1931-1947
[7] Fu Zhou-yu,Robles-Kelly A,Zhou Jun.MILIS:multiple in-stance learning with instance selection [J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2011,33(5):958-977
[8] Zhang Gang,Yin Jian,Cheng Liang-lun,et al.Semi-supervisedmulti-instance kernel [J].Computer Science,2011,38(9):220-223(in Chinese) 张钢,印鉴,程良伦,等.半监督多示例核 [J].计算机科学,2011,38(9):220-223
[9] Grtner T,Flach P A,Kowalczyk A,et al.Multi-instance kernels[C]∥Proceedings of the 19th International Conference on Machine Learning,2002.Sydney,Australia,2002:179-186
[10] Li Bing,Xiong Wei-hua,Hu Wei-ming.Context-aware multi-instance learning based on hierarchical sparse representation[C]∥IEEE 11th International Conference on Data Mining,2011.Vancouver,Canada:IEEE,2011:370-377
[11] Liu Guo-qing,Wu Jian-xin,Zhou Zhi-hua.Key instance detection in multi-instance learning[C]∥Proceedings of the 4th Asian Conference on Machine Learning,2012.Singapore,2012:253-268
[12] Ding Xin-miao,Li Bing,Hu Wei-ming,et al.Horror video scene recognition based on multi-view joint sparse coding [J].Acta Electronica Sinica,2014,42(2):301-305(in Chinese) 丁昕苗,李兵,胡卫明,等.基于多视角融合稀疏表示的恐怖视频识别 [J].电子学报,2014,42(2):301-305
[13] Feng Song-he,Xiong Wei-hua,Li Bing,et al.Hierarchical sparse representation based multi-instance semi-supervised learning with application to image categorization [J].Signal Processing,2014,94(1):595-607
[14] Zhou Zhi-hua,Zhang Min-ling.Multi-instance multi-label lear-ning with application to scene classification[C]∥Advances in Neural Information Processing Systems 19,2007.Cambridge,United Kingdom:MIT Press,2007:1609-1616
[15] Zhou Zhi-hua,Zhang Min-ling,Huang Sheng-jun,et al.Multi-instance multi-label learning [J].Artificial Intelligence,2012,176(1):2291-2320
[16] Li Yu-feng,Hu Ju-huan,Jiang Yuan,et al.Towards discovering what patterns trigger what labels[C]∥Proceedings of the 26th AAAI Conference on Artificial Intelligence,2012.Toronto,Ca-nada,2012:1012-1018
[17] Yang Shu-jun,Jiang Yuan,Zhou Zhi-hua.Multi-instance multi-label learning with weak label[C]∥Proceedings of the 23rd International Joint Conference on Artificial Intelligence,2013.Beijing,China:AAAI Press,2013:1862-1868
[18] Huang Sheng-jun,Gao Wei,Zhou Zhi-hua.Fast multi-instance multi-label learning[C]∥Proceedings of the 28th AAAI Confe-rence on Artificial Intelligence,2014.Quebec City,Canada,2014:1868-1874
[19] Zhou Zhi-hua,Sun Yu-yin,Li Yu-feng.Multi-instance learningby treating instances as non-I.I.D.samples[C]∥Proceedings of the 26th International Conference on Machine Learning,2009.Montreal,Canada,2009:1249-1256
[20] Rakotomamonjy A,Bach F R,Canu S,et al.SimpleMKL [J].Journal of Machine Learning,2008,9(11):2491-2521
[21] Briggs F,Fern X,Raich R.Rank-loss support instance machines for miml instance annotation[C]∥Proceedings of the 18th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining,2012.Beijing,China,2012:534-542
[22] Maron O,Ratan A L.Multiple-instance learning for naturalscene classification[C]∥Proceedings of the 15th International Conference on Machine Learning,1998.San Francisco,America,1998:341-349
[23] Andrews S,Tsochantaridis I,Hofmann T.Support vector machines for multiple-instance learning[C]∥Advances in Neural Information Processing Systems 15,2003.Cambridge,United Kingdom:MIT Press,2003:561-568
[24] Zhang Min-ling,Zhou Zhi-hua.Multi-label learning by instance differentiation[C]∥roceedings of the 22nd AAAI Conference on Artificial Intelligence.Vancouver,Canada:AAAI Press,2007:669-674
[25] Zhang Min-ling,Zhou Zhi-hua.Multi-label learning by instance differentiation[J].Pattern Recognition,2007,40(7):2038-2048
[26] Hao Hong,Ji Hua,Zhang Hua-xiang,et al.Multi-label sceneclassification based on I2C distance and label dependency [J].Computer Science,2014,41(1):88-90(in Chinese) 郝虹,计华,张化祥,等.基于I2C距离和标记相关性的多标记场景分类 [J].计算机科学,2014,41(1):88-90

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed