Graph-Margin Based Multi-Label Feature Selection

Abstract

Since instances in multi-label problems are associated with several labels simultaneously, most traditional feature selection algorithms for single label problems are inapplicable. Therefore, new criteria to evaluate features and new methods to model label correlations are needed. In this paper, we adopt the graph model to capture the label correlation, and propose a feature selection algorithm for multi-label problems according to the graph combining with the large margin theory. The proposed multi-label feature selection algorithm GMBA can efficiently utilize the high order label correlation. Experiments on real world data sets demonstrate the effectiveness of the proposed method. The codes of the experiment of this paper are available at https://github.com/Faustus-/ECML2016-GMBA .

Cite

Text

Yan and Li. "Graph-Margin Based Multi-Label Feature Selection." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2016. doi:10.1007/978-3-319-46128-1_34

Markdown

[Yan and Li. "Graph-Margin Based Multi-Label Feature Selection." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2016.](https://mlanthology.org/ecmlpkdd/2016/yan2016ecmlpkdd-graphmargin/) doi:10.1007/978-3-319-46128-1_34

BibTeX

@inproceedings{yan2016ecmlpkdd-graphmargin,
  title     = {{Graph-Margin Based Multi-Label Feature Selection}},
  author    = {Yan, Peng and Li, Yun},
  booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
  year      = {2016},
  pages     = {540-555},
  doi       = {10.1007/978-3-319-46128-1_34},
  url       = {https://mlanthology.org/ecmlpkdd/2016/yan2016ecmlpkdd-graphmargin/}
}