摘要
退化方法是求解多示例多标记学习(MIML)问题常用的求解方式,但是在退化过程中会造成标记之间的关联信息丢失。对该问题进行研究,提出OCC-MIMLSVM+分类算法,将MIMLSVM+算法与有序分类器链(OCC)方法相结合,通过对分类器进行合理组织,将标记之间的关联信息融入至算法的训练过程中,解决信息丢失问题,提高分类准确率。实验结果表明,改进算法取得了比基准多示例多标记算法更好的分类效果。
Degradation is a common solution to the problem of the MIML classification problem.However,the correlation information among labels may lost in the degradation process.Based on these problems,the OCC-MIMLSVM+algorithm was proposed.The MIMLSVM+algorithm was combined with the ordered classifier chain(OCC)method,the classifier was organized and the dependency relation between labels was integrated into the training process of the algorithm,so that the problem of information lost was solved,and the accuracy of classification was improved.Experimental results show that the improved algorithm achieves better classification results than the benchmark multi-instance multi-label algorithm.
引文
[1]Zhou Z H.A framework for machine learning with ambiguous objects[M]//Brain Informatics.Springer Berlin Heidelberg,2009:6.
[2]Wei X S,Wu J,Zhou Z H.Scalable algorithms for multi-instance learning[J].IEEE Transactions on Neural Networks and Learning Systems,2017,28(4):975-987.
[3]Wang B,Tsotsos J.Dynamic label propagation for semi-supervised multi-class multi-label classification[J].Pattern Recognition,2016,52(C):75-84.
[4]Keikha M,Hashemi S.Ordered classifier chains for multi-label classification[J].Journal of Machine Intelligence,2016,1(1):7-12.
[5]Zhou Z H,Zhang M L,Huang S J,et al.Multi-instance multi-label learning[J].Artificial Intelligence,2008,176(1):2291-2320.
[6]Li Y X,Ji S,Kumar S,et al.Drosophila gene expression pattern annotation through multi-instance multi-label learning[J].IEEE/ACM Transactions on Computational Biology&Bioinformatics,2012,9(1):98.
[7]Zhang M L,Zhou Z H.M3MIML:A maximum margin method for multi-instance multi-label learning[C]//Eighth IEEEInternational Conference on Data Mining.IEEE,2008:688-697.
[8]Li Y,Wang S,Tian Q,et al.A boosting approach to exploit instance correlations for multi-instance classification[J].IEEETransactions on Neural Networks&Learning Systems,2016,27(12):2740-2747.
[9]Peker M.A decision support system to improve medical diagnosis using a combination of k-medoids clustering based attribute weighting and SVM[J].Journal of Medical Systems,2016,40(5):1-16.
[10]Su C,Yang F,Zhang S,et al.Multi-task learning with low rank attribute embedding for person re-identification[C]//IEEE International Conference on Computer Vision.IEEE,2015:3739-3747.
[11]Babcock B,Babu S,Datar M,et al.Learning multiple tasks with kernel methods[J].Journal of Machine Learning Research,2014,6(4):615-637.
[12]Briggs F,Fern X Z,Raich R.Context-aware MIML instance annotation:Exploiting label correlations with classifier chains[J].Knowledge&Information Systems,2014,43(1):1-27.
[13]Li C,Zhang Y,Lu L.An MIMLSVM algorithm based on ECC[J].Applied Intelligence,2015,42(3):537-543.
[14]Huang J,Li G,Wang S,et al.Group sensitive classifier chains for multi-label classification[C]//IEEE International Conference on Multimedia and Expo.IEEE Computer Society,2015:1-6.
[15]Andrews S,Tsochantaridis I,Hofmann T.Support vector machines for multiple-instance learning[M].HCI International 2016-Posters’Extended Abstracts.Springer International Publishing,2016.