Abstract In online classification tasks, new class of patterns sometimes emerges, which makes the distribution change significantly and current classification models invalid. A method based on distance metric learning is proposed to recognize new class from existing classes without the apriori knowledge about emerging class. And it can make the class similarity represented by using distance between two objects, which is the key of promoting the performance of recognition. Therefore, the proposed method can be applied to the adaptive classification. The experimental results show that the proposed method can recognize new class well, and on the basis of this method the online classifier is adapted and it can predict the instance better than the original one.
[1] Cuevas A, Febrero M, Fraiman R. Cluster Analysis: A Further Approach Based on Density Estimation. Computational Statistics and Data Analysis, 2001, 36(4): 441-459 [2] Chang I, Loew M. Pattern Recognition with New Class Discovery // Proc of the IEEE International Conference on Computer Vision and Pattern Recognition. Maui, USA, 1991: 438-443 [3] Chow C. Parzen-Window Network Intrusion Detectors // Proc of the 16th International Conference on Pattern Recognition. Québec, Canada, 2002: 385-388 [4] Lauer M. A Mixture Approach to Novelty Detection Using Training Data with Outliers // Proc of the 12th European Conference on Machine Learning. Freiburg, Germany, 2001: 300-311 [5] Liu Yan, Gururajan S, Cukic B, et al. Validating an Online Adaptive System Using SVDD // Proc of the 15th International Conference on Tools with Artificial Intelligence. Sacramento, USA, 2003: 384-388 [6] Tax D M J, Duin R P W. Support Vector Data Description. Machine Learning, 2004, 54(1): 45-66 [7] Ougiaroglou S, Nanopoulos A, Papadopoulos A N, et al. Adaptive k-Nearest-Neighbor Classification Using a Dynamic Number of Nearest Neighbors // Proc of the 11th East-European Conference on Advances in Databases and Information Systems. Varna, Bulgaria, 2007: 66-82 [8] Krishnapuram B. Adaptive Classifier Design Using Labeled and Unlabeled Data. Ph.D Dissertation. Durham, USA: Duke University. Department of ECE, 2004 [9] Zhou Dengyong, Bousquet O, Lal T N, et al. Learning with Local and Global Consistency // Thrun S, Saul L K, Scholkpf B, eds. Advances in Neural Information Processing System. Cambridge, USA: MIT Press, 2004, 16: 321-328 [10] Xie Maoqiang, Huang Yalou. Adaptive Algorithm for Class Incremental Induction of Decision Tree. Computer Engineering, 2006, 32(17): 41-43 (in Chinese) (谢茂强,黄亚楼.适应类别增量的决策树训练算法.计算机工程, 2006, 32(17): 41-43) [11] Song Rui, Zhang Jing, Xia Shengpin, et al. An Adaptive Classification Method of BP-NN Group Based Classification System and Its Application. Acta Electronica Sinica, 2001, 29(12): 1950-1953 (in Chinese) (宋 锐,张 静,夏胜平,等.一种基于BP 神经网络群的自适应分类方法及其应用.电子学报, 2001, 29(12): 1950-1953) [12] Xing E P, Ng A Y, Jordan M I, et al. Distance Metric Learning, with Application to Clustering with Side-Information // Becker S, Thrun S, Obermayer K, eds. Advances in Neural Information Processing System. Cambridge, USA: MIT Press, 2003, 15: 505-512
2009, Vol. 22 
