Stable L2-Regularized Ensemble Feature Weighting
When selecting features for knowledge discovery applications, stability is a highly desired property. By stability of feature selection, here it means that the feature selection outcomes vary only insignificantly if the respective data change slightly. Several stable feature selection methods have been proposed, but only with empirical evaluation of the stability. In this paper, we aim at providing a try to give an analysis for the stability of our ensemble feature weighting algorithm. As an example, a feature weighting method based on L2-regularized logistic loss and its ensembles using linear aggregation is introduced. Moreover, the detailed analysis for uniform stability and rotation invariance of the ensemble feature weighting method is presented. Additionally, some experiments were conducted using real-world microarray data sets. Results show that the proposed ensemble feature weighting methods preserved stability property while performing satisfactory classification. In most cases, at least one of them actually provided better or similar tradeoff between stability and classification when compared with other methods designed for boosting the stability.
Unable to display preview. Download preview PDF.
- 1.Ng, A.Y.: Feature selection, l1 vs. l2 regularization, and rotational invariance. In: Proceedings of International Conference on Machine Learning, Banff, Canada (2004)Google Scholar
- 2.Zhao, Z.: Spectral Feature Selection for Mining Ultrahigh Dimensional Data. PhD thesis, Arizona State University (2010)Google Scholar
- 7.Guyon, I., Elisseeff, A.: An introduction to variable and feature selection. Journal of Machine Learning Research 31, 1157–1182 (2003)Google Scholar
- 9.Han, Y., Yu, L.: A variance reduction for stable feature selection. In: Proceedings of the International Conference on Data Mining, pp. 206–215 (2010)Google Scholar
- 10.Loscalzo, S., Yu, L., Ding, C.: Consensus group stable feature selection. In: Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 567–575 (2009)Google Scholar
- 15.Li, Y., Gao, S.Y., Chen, S.C.: Ensemble feature weighting based on local learning and diversity. In: AAAI Conference on Artificial Intelligence, pp. 1019–1025 (2012)Google Scholar
- 16.Woznica, A., Nguyen, P., Kalousis, A.: Model mining for robust feature selection. In: Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp. 913–921 (2012)Google Scholar
- 19.Crammer, K., Bachrach, R.G., Navot, A., Tishby, N.: Margin analysis of the lvq algorithm. In: Advances in Neural Information Processing Systems, pp. 462–469 (2002)Google Scholar
- 22.Breiman, L.: Bagging predictors. Machine Learning 26, 123–140 (1996)Google Scholar
- 25.Frank, A., Asuncion, A.: UCI machine learning repository (2010), http://archive.ics.uci.edu/ml
- 26.Alon, U., Barkai, N., Notterman, D.A., Gish, K., Ybarra, S., Mack, D., Levine, A.J.: Broad patterns of gene expression revealed by clustering analysis of tumor and normal colon cancer tissues probed by oligonucleotide arrays. Proceedings of the National Academy of Sciences of the United States of America, 6745–6750 (1999)Google Scholar
- 27.Bhattacharjee, A., Richards, W.G., Staunton, J., Li, C., Monti, S.: Classification of human lung carcinomas by mrna expression profiling reveals distinct adenocarcinoma subclasses. Proceedings of the National Academy of Sciences of the United States of America 98, 13790–13795 (2001)CrossRefGoogle Scholar